Video Large Language Models (Video-LLMs) have made rapid progress on temporal video understanding, yet many fail at a basic perceptual primitive: signed image-plane motion direction. On simple videos of a single object moving left, right, up, or down, most Video-LLMs perform near chance, with above-chance cases largely attributable to prediction biases rather than genuine direction understanding. We call this failure directional motion blindness. We localize the failure by tracing motion direction information through the Video-LLM pipeline. Motion direction remains linearly accessible from the vision encoder, projector, and LLM hidden states, but the readout fails to bind this signal to the correct verbal answer option, revealing a direction binding gap. Although synthetic motion direction instruction tuning reduces this gap on the source domain, motion direction concept vector analysis shows that visual complexity weakens the signal magnitude and limits out-of-domain generalization. We introduce MoDirect, a dataset family for motion direction instruction tuning and evaluation, and DeltaDirect, a diagnosis-driven, projector-level objective that predicts normalized 2-D motion vectors from adjacent-frame feature deltas. On MoDirect-SynBench, instruction tuning with DeltaDirect improves motion direction accuracy from 25.9% to 85.4%. On MoDirect-RealBench, DeltaDirect improves real-world motion direction accuracy by 21.9 points over the vanilla baseline without real-world tuning data, while preserving standard video-understanding performance. Code: https://github.com/KHU-VLL/DeltaDirect

Camera pose matters. The position and orientation of each viewpoint define a shared spatial coordinate frame that relates observations across video frames. Yet this signal is largely absent from multimodal LLMs (MLLMs) for video understanding, which process frames as isolated 2D snapshots, instead of the persistent scene humans perceive. We revisit pose as a lightweight supervisory signal and introduce Cambrian-P, a video MLLM augmented with per-frame learnable camera tokens and a pose regression head. With a carefully designed sampling scheme, the model achieves substantial gains of 4.5-6.5% on spatial reasoning benchmarks such as VSI-Bench, generalizes across eight additional spatial and general video QA benchmarks, and, as a byproduct, achieves state of the art streaming pose estimation on ScanNet. Surprisingly, training on pseudo-annotated poses from in-the-wild video further improves general video QA benchmarks, showing pose helps beyond spatial reasoning. Together, these results position camera pose as a fundamental signal for video models that reason about the physical world.

Current motion-controlled image-to-video generation models rigidly follow user-provided trajectories that are often sparse, imprecise, and causally incomplete. Such reliance often yields unnatural or implausible outcomes, especially by missing secondary causal consequences. To address this, we introduce MotiMotion, a novel framework that reformulates motion control as a reasoning-then-generation problem. To encourage causally grounded and commonsense-consistent interactions, we leverage a training-free vision-language reasoner to refine image-space coordinates of primary trajectories and to hallucinate plausible secondary motions. To further improve motion naturalness, we propose a confidence-aware control scheme that modulates guidance strength, enabling the model to closely follow high-confidence plans while correcting artifacts under low-confidence inputs with its internal generative priors. To support systematic evaluation, we curate a new image-to-video benchmark, MotiBench, consisting of interaction-centric scenes where new events are triggered by motion. Both VLM-based evaluation and a human study on MotiBench demonstrate that MotiMotion produces videos with more plausible object behaviors and interaction, and is preferred over existing approaches.

Vision-and-Language Navigation (VLN) requires an agent to ground language instructions to its own movement within a visual environment. While state-of-the-art methods leverage the reasoning capabilities of Vision-Language Models (VLMs) for end-to-end action prediction, they often lack an explicit and explainable understanding of the relationships between the agent, the instruction, and the scene. Conversely, explicitly building a scene map for heuristic planning is intuitively appealing but relies on additional 3D sensors and hinders large-scale vision-language pre-training. To bridge this gap, we propose AwareVLN, a novel framework that equips the navigation model with a self-aware reasoning mechanism, enabling it to understand the agent's state and task progress in a fully end-to-end and data-driven manner. Our approach features two key innovations: (1) a structural reasoning module that fosters spatial and task-oriented self-awareness, and (2) an automatic data engine with progress division for effective training. Extensive experiments on various datasets in Habitat simulator show our AwareVLN significantly outperforms previous state-of-the-art vision-language navigation methods. Project page: https://gwxuan.github.io/AwareVLN/.

Vision-Language-Action (VLA) models have shown strong potential for general-purpose robot manipulation by unifying perception and action. However, existing VLA systems primarily rely on textual instructions and struggle to resolve spatial ambiguity in complex scenes with multiple similar objects. To address this limitation, we introduce gesture as a parallel instruction modality and propose a Gesture-aware Vision-Language-Action model (GesVLA). Our approach encodes gesture features directly into the latent space, enabling them to participate in both high-level reasoning and low-level action generation, and adopts a dual-VLM architecture to achieve tight coupling between gesture representations and action policies. At the data level, we construct a scalable gesture data generation pipeline by rendering hand models onto real-world scene images. This reduces the sim-to-real visual gap while producing rich data with diverse motion patterns and corresponding pointing annotations. In addition, we employ a two-stage training strategy to equip the model with both gesture perception and action prediction capabilities. We evaluate our approach on multiple real-world robotic tasks, including a controlled block manipulation task for validation and more practical scenarios such as product and produce selection. Experimental results show that incorporating gesture consistently improves target grounding accuracy and human-robot interaction efficiency, especially in complex and cluttered environments. Project page: https://gwxuan.github.io/GesVLA/.

Representation Autoencoders (RAEs) leverage frozen vision foundation models (VFMs) as tokenizer encoders, providing robust high-level representations that facilitate fast convergence and high-quality generation in latent diffusion models. However, freezing the VFM inherently constrains its spatial reconstruction capacity, limiting fine-grained generation and image editing; in contrast, incorporating reconstruction-oriented signals via fine-tuning disrupts the pretrained semantic space and degrades generative fidelity. To address this trade-off, we propose DecQ, a simple yet effective framework for RAEs. Specifically, DecQ introduces lightweight detail-condensing queries that extract fine-grained information from intermediate VFM features through condenser modules. These queries are incorporated into the decoder to support reconstruction and are jointly generated with patch tokens during generative modeling. By aggregating information from both shallow and deep layers, DecQ effectively mitigates the reconstruction--generation trade-off, improving both reconstruction quality and generative performance. Our experiments demonstrate that: (1) with only 8 additional queries and 3.9% extra computation, DecQ improves reconstruction over the frozen DINOv2-based RAE, increasing PSNR from 19.13 dB to 22.76 dB; and (2) for generative modeling, DecQ achieves 3.3$\times$ faster convergence than RAE, attaining an FID of 1.41 without guidance and 1.05 with guidance.

Children with rare genetic diseases often exhibit distinctive facial phenotypes, yet developing computer vision systems for early diagnosis remains challenging due to extreme data scarcity, privacy constraints, and limited data sharing in pediatric settings. These challenges not only hinder automated diagnosis but also restrict the availability of visual resources for clinical genetic counseling. While prior work has shown that synthetic data can augment real datasets and preserve phenotype-level semantics, it remains unclear whether synthetic data alone is sufficient for learning in ultra-low-resource pediatric settings. In this work, we study the synthetic-only regime for pediatric rare disease recognition. Under a controlled experimental setup, models are trained exclusively on phenotype-aware synthetic facial images at increasing scales. We find that synthetic-only training achieves performance comparable to real-data-only baselines at sufficient scale across multiple backbones, suggesting that high-fidelity synthetic data can approximate clinically meaningful distributions. These findings together further enable the use of synthetic pediatric facial images as privacy-preserving resources for genetic education and counseling, supporting clinician training and patient communication. Our results highlight the potential of computer vision to improve data efficiency and expand accessible visual tools in children's healthcare.

As generative image models evolve rapidly, the perceptual gap between generated and real images continues to narrow, making AI-generated image detection increasingly challenging. Many existing methods exploit frequency-domain cues for detection, typically described as frequency-domain artifacts or high-frequency discrepancies. However, the specific and recurring spectral regularities remain insufficiently understood and characterized. In this paper, we systematically analyze the one-dimensional radial log-power spectra of real and generated images. We find that generated images do not necessarily exhibit higher or lower energy across the entire spectrum or high-band range. Instead, their spectra deviate from the power-law decay and show an anomalous uplift in the ultra-high-frequency tail. We term this phenomenon spectral tail uplift. We further attribute this phenomenon to nonlinear harmonic accumulation in trained generative models, suggesting that it can serve as a structural cue across generative architectures. Based on this observation, we propose Spectral Tail Auxiliary Learning (STAL), a frequency-domain auxiliary supervision framework for generalizable AI-generated image detection. STAL transfers spectral-tail cues from a tail-aware frequency teacher to a spatial detector during training, while all frequency-domain modules are discarded at inference time. Consequently, STAL introduces no inference overhead. Extensive experiments on 9 public datasets show that STAL achieves strong generalization and stability across generators, data distributions, and real-world scenarios.

Autoregressive video diffusion models have enabled real-time, action-conditioned world generation. However, sustaining a persistent world, where revisiting a previously seen viewpoint yields consistent content, remains an open problem. Full KV-cache attention preserves this consistency but breaks real-time constraints: memory footprint and attention cost grow linearly with rollout length. Sliding window inference restores throughput but discards long-term consistency. We propose WorldKV, a training-free framework with two components: World Retrieval and World Compression. World Retrieval stores evicted KV-cache chunks in GPU/CPU memory and selectively retrieves scene-relevant chunks via camera/ action correspondence, inserting them back into the native attention window without re-encoding. World Compression prunes redundant tokens within each chunk via key-key similarity to an anchor frame, halving per-chunk storage to fit 2x more history under a fixed budget. On Matrix-Game-2.0 and LingBot- World-Fast, WorldKV matches or exceeds full-KV memory fidelity at roughly 2x the throughput, and is competitive with memory-trained baselines without any fine-tuning. Project Page: https://cvlab-kaist.github.io/WorldKV/

Robustness to domain changes is a key capability for effective deployment of human action recognition systems in real-world scenarios, where action categories at inference can present important domain shifts or even unseen actions from training. In this context, improving the recognition capabilities of Zero-Shot Action Recognition models (ZSAR), without requiring strong annotation efforts, remains a central challenge. Most ZSAR approaches assume that actions are observed under geometric conditions similar to those seen during training. In practice, variations in human body orientation and camera viewpoint add a significant domain gap in ZSAR, substantially limiting generalization to novel action-motion combinations. In this context, this paper presents a novel orientation-aware action recognition approach with improved cross-domain capabilities. Our approach combines motion cues of multiple camera viewpoints and text descriptions of human actions in the training phase. We present a new orientation-aware motion encoding network to learn different motion features, and adapt a specific orientation-aware text prompt to match the corresponding features at inference. Extensive experiments demonstrate that the proposed method consistently improves ZSAR performance across different recognition benchmarks, outperforming recent state-of-the-art zero-shot approaches on NTU-RGB+D, BABEL, NW-UCLA, and on two surveillance datasets. In addition, the learned representations exhibit strong transfer learning capabilities, yielding competitive performance on both cross-domain and same-domain recognition of seen actions. Code and trained models are available at: https://icb-vision-ai.github.io/OrientationAware-HAR

Viewpoint change invariance and action temporal consistency are critical aspects for the effective deployment of human action detection of untrimmed videos. Existing appearance-based video detection methods often struggle with limited viewpoint diversity during training, while motion-based detection approaches frequently fail to model fine-grained temporal relationships across consecutive motion windows. This paper introduces a novel two-stage action detection approach designed to improve both view-invariance and global temporal coherence properties. In the first stage, we extract motion features from augmented virtual viewpoints, solely used at training. Then, the second stage introduces a new view-invariant, multi-scale temporal encoder based on selective state-space sequence modelling to aggregate information across viewpoints and time scales. Experiments on PKU-MMD and BABEL benchmarks demonstrate that this approach significantly outperforms state-of-the-art methods in all considered splits. Code and trained models are available at: https://icb-vision-ai.github.io/HydraView-TAD

Vision-language models learn powerful multimodal embeddings, yet their internal semantics remain opaque. While sparse autoencoders (SAEs) can extract interpretable features, they rely on expanding the representation dimension, which compromises the original geometry and introduces redundancy. We introduce CEDAR (Conceptual Embedding Disentanglement via Adaptive Rotation), a post-hoc method that reveals the compositional structure of pretrained embeddings without increasing dimensionality. By learning an invertible transformation with a top-$k$ sparsity bottleneck, CEDAR concentrates semantic information into axis-aligned disentangled coordinates. In CLIP-like architecture, individual coordinates can be interpreted with textual concepts, while for generative models such as BLIP, they can be decoded into natural language descriptions. Experiments demonstrate that CEDAR achieves a competitive reconstruction-sparsity trade-off while producing explanations that are more interpretable and better aligned with human perception. Our results suggest that the apparent entanglement in vision-language representations can be resolved through a suitable change of basis, eliminating the need for overcomplete expansions.

While most frames in long-form video are redundant, the critical information resides in temporal surprises: moments where the actual visual features deviate from their predicted evolution. Inspired by the human brain's predictive coding, we introduce Swift Sampling, an elegant, training-free frame selection algorithm that automatically identifies high-information moments in a video. Specifically, we model a video as a differentiable trajectory in the visual latent space and compute the velocity and acceleration of its features. Then, we apply Taylor expansion to project the expected path of subsequent frames. Frames that diverge sharply from this predicted manifold are identified as temporally surprising frames and selected for sampling. Unlike prior training-free methods that rely on auxiliary networks or video-specific hyperparameter tuning, Swift Sampling is incredibly lightweight, adding only 0.02x additional computational cost over baseline making it 30x cheaper overhead than leading baselines. Across three long-video question answering benchmarks and 10 different downstream tasks, Swift Sampling outperforms uniform sampling and prior query-agnostic baselines. It is especially powerful for long videos with limited frame budgets improving accuracy by up to +12.5 points.

Deploying vision models across devices with varying resource constraints, or even on a single device where available compute fluctuates due to battery state, thermal throttling, or latency deadlines, typically requires training and maintaining separate models. Width-adaptive inference addresses this by training a single set of shared weights containing multiple nested subnetworks of increasing capacity, but prior CNN-based approaches required switchable batch normalization, while recent scalable methods have focused on Vision Transformers. We present Slimmable ConvNeXt, which shows that ConvNeXt's modern design, specifically LayerNorm and inverted bottlenecks, makes it particularly suited for channel-width slimming, eliminating the normalization overhead of classical slimmable networks and producing a simpler training pipeline than both prior CNN and ViT approaches. On ImageNet-1k, Slimmable ConvNeXt-T with 3 subnetworks achieves 80.8% top-1 accuracy at 4.5 GMACs and 77.4% at 1.2 GMACs, trained from scratch for 600 epochs. At comparable compute, this exceeds HydraViT's 6-head subnetwork (78.4% at 4.6 GMACs) by 2.4 percentage points and its 3-head configuration (73.0% at 1.3 GMACs) by 4.4 percentage points, while also outperforming MatFormer-S (78.6%) and SortedNet-S (78.2%) at the same GMACs. Scaling to Slimmable ConvNeXt-B further improves maximum accuracy to 82.8% at 15.35 GMACs.

Diffusion transformers (DiTs) have emerged as a dominant architecture for text-to-image generation, yet their performance drops when generating at resolutions beyond their training range. Existing training-free approaches mitigate this by modifying inference-time attention behavior, often through Rotary Position Embeddings (RoPE) extrapolation combined with attention scaling. However, these strategies apply a uniform and content-agnostic scaling across RoPE components with distinct frequency characteristics, inducing a trade-off between preserving global structure and recovering fine detail. We introduce SEGA, a training-free method that dynamically scales attention across RoPE components according to the latent's spatial-frequency structure at each denoising step. This adaptive scaling improves both structural coherence and fine-detail fidelity. Experiments show that SEGA consistently improves high-resolution synthesis across multiple target resolutions, outperforming state-of-the-art training-free baselines.

While large language models provide strong compositional reasoning, existing reasoning segmentation pipelines fail to transparently connect this reasoning to visual perception. Current methods, such as latent query alignment, are end-to-end yet opaque "black boxes". Conversely, textual localization readout is merely readable, not truly interpretable, often functioning as an unconstrained post-hoc step. To bridge this interpretability gap, we propose SegCompass, an end-to-end model that leverages a Sparse Autoencoder (SAE) to forge an explicit, interpretable, and differentiable alignment pathway. Given an image-instruction pair, SegCompass first generates a chain-of-thought (CoT) trace. The core of our method is an SAE that maps both the CoT and visual tokens into a shared, high-dimensional sparse concept space. A query codebook selects salient concepts from this space, which are then spatially grounded by a slot mapper into a multi-slot heatmap that guides the final mask decoder. The entire model is trained jointly, unifying reinforcement learning for the reasoning path with standard segmentation supervision. This SAE-driven interface provides a "white-box" connection that is significantly more traceable than latent queries and more coherent than textual readouts. Extensive experiments on five challenging benchmarks demonstrate that SegCompass matches or surpasses state-of-the-art performance. Crucially, our visual and quantitative analyses show a strong correlation between the quality of the learned sparse concepts and final mask accuracy, confirming that SegCompass achieves superior results through its enhanced and inspectable alignment. Code is available at https://github.com/ZhenyuLU-Heliodore/SegCompass.

Previous detection studies have shown that LLMs cannot be effectively used as detectors, but these studies have not addressed modern Chinese poetry. Moreover, no relevant research has explored the performance of LLMs in detecting modern Chinese poetry. This paper evaluates and enhances the performance of LLMs as detectors for modern Chinese poetry, and proposes an image-semantic guided poetry detection method. Compared with traditional detection approaches, our method innovatively incorporates images that reflect the content of the poetry. Through example-driven approaches, our method effectively integrates information such as meaning, imagery, and feeling from the image, then forms a complementary judgment with the poem text. Experimental results demonstrate that the LLM detectors based on our method outperform baseline detectors based on plain text, and even surpass the best-performing traditional detector, RoBERTa. The Gemini detector using our method achieves a Macro-F1 score of 85.65%, reaching the state-of-the-art level. The performance improvements of different LLM detectors on multiple LLMs-generated data prove the effectiveness of our method.

CLIP-style contrastive pretraining typically curates web-scale image-text pairs using sample-level filtering signals, often based on pair-level alignment. We show that this signal saturates: once coarse mismatches are removed, stricter global filtering no longer tracks the compositional supervision provided by the retained captions. The reason is structural - a global score conflates whether a pair is broadly plausible with whether the individual object, attribute, and relation phrases inside the caption materially support the image-text match. The latter is what compositional generalization demands, yet pair-level filters are blind to it. We address this with Counterfactual Phrase Intervention (CPI), a phrase-level curation framework that converts controlled nonce-token substitutions into image-conditioned phrase-sensitivity scores. CPI uses global alignment only for coarse mismatch removal, then ranks the surviving pool by whether caption phrases measurably affect the image-text score under controlled substitution. We frame CPI as a first-order phrase-sensitivity signal rather than a grounding or identification result, and evaluate it at CC3M scale. Ranking by this signal yields a 50%-data subset that improves VL-CheckList-VG Relation by +1.91 over the full-data baseline and +1.00 over alignment-only filtering at matched budget, while improving SugarCrepe overall and preserving general transfer. CPI is loss-orthogonal: applied unchanged to NegCLIP, it further improves VL-CheckList-VG Relation by +3.84, with additional CE-CLIP gains in the main text.

Dual-energy X-ray absorptiometry (DXA) is widely used for large-scale skeletal assessment, yet learning controllable and interpretable factor-specific anatomical variation remains challenging. We propose a metadata-conditioned causal hierarchical variational autoencoder (CHVAE) for causally consistent generation of anteroposterior (AP) spine DXA images from the UK Biobank (UKB). The model is trained on 3,743 raw AP spine scans from the first imaging visit and conditioned on basic participant attributes and lumbar morphometry. Causal consistency is evaluated in a baseline-to-follow-up setting using abduction--action--prediction (AAP): latent variables are abducted from baseline images, age is intervened to the repeat-imaging value, and the resulting counterfactual follow-up morphometry is compared with observed repeat-imaging measurements. Results show strong absolute-level agreement for key vertebral morphometry variables under age intervention, supporting intervention-aligned synthesis of anatomically plausible DXA images.

Accurately reconstructing full-body poses from sparse head and hand trajectories is a foundational challenge for immersive AR/VR telepresence. Current methods often struggle with error accumulation and unnatural joint coordination, primarily because they treat the human body as a monolithic entity, thereby failing to capture the fine-grained ``atomic intents'' embedded in subtle signal variations and overlooking the inherent structural topology. To bridge this gap, we present AtomicMotion, a framework designed to decouple and re-integrate body dynamics through three core innovations. First, we introduce a logical body partitioning scheme that decomposes the skeleton into five distinct clusters based on functional intent; this ensures that each partition preserves internal joint synergies while isolating local motion primitives. Second, to robustly map sparse inputs to high-dimensional poses, we employ a masked full-body pre-conditioning strategy during training, forcing the model to internalize global skeletal topology and latent kinematic constraints. Finally, addressing the limitations of vanilla spatial attention, which often ignores fixed physiological connectivity, we propose Kinematic Attention. By embedding the classical kinematic tree structure into the attention mechanism, we ensure biological plausibility in the synthesized motions. Extensive evaluations on the AMASS dataset demonstrate that AtomicMotion significantly outperforms existing baselines, yielding higher reconstruction fidelity and superior biomechanical realism.

Parametric human models capture global pose but cannot represent the non-rigid surface dynamics of clothing and soft tissue. Generic scene flow estimates dense motion but breaks down on articulated bodies, where pixel-level supervision is also intractable to acquire. We introduce H-Flow, a dense human scene flow that captures both skeletal kinematics and surface deformation. A unified multi-head transformer estimates flow from monocular video, jointly predicting pose and depth as companion outputs. The challenge lies in the lack of supervision. In place of unattainable labels, we anchor the network in the physics of human motion, encoding geometric, structural, and biomechanical priors as cross-modal training objectives. We further introduce DynAct4D, a high-fidelity synthetic benchmark providing dense flow annotations across diverse subjects, garments, and motions. On standard benchmarks, H-Flow outperforms scene-flow and parametric baselines, and generalizes zero-shot to in-the-wild video. Code, models, and the DynAct4D benchmark will be released upon publication

Grounding radiology report descriptions to 3D CT volumes is essential for verifiable clinical interpretation, yet remains challenging due to the semantic-spatial gap between free-text narratives and volumetric anatomy. Existing report-assisted and vision-language grounding methods typically rely on phrase-level alignment or dense pixel supervision, resulting in limited lesion-wise correspondence and suboptimal localization accuracy. We propose GLeVE, a graph-guided lesion grounding framework with anatomical prior verification and octree-based autoregressive refinement. GLeVE treats each lesion description as an atomic semantic unit and encodes organ attribution, attributes, and inter-lesion relations through relation-aware graph reasoning to produce discriminative lesion-wise queries. Anatomy-aware proposal generation with region-level verification enforces one-to-one text-lesion alignment, while hierarchical octree refinement progressively improves boundary delineation. Experiments on AbdomenAtlas 3.0 demonstrate consistent gains over classical multimodal foundation models and report-supervised baselines in both segmentation accuracy and lesion-level localization.

Gaze following requires both scene understanding and gaze reasoning to localize the gaze target of an in-scene person. Recently, vision foundation models (VFMs) have demonstrated strong performance on this task, enabling simpler architectures while outperforming prior methods. However, we observe a key limitation of VFM-based approaches: while VFMs substantially improve scene understanding, they contribute little to gaze reasoning. As a result, existing methods often rely on semantically salient objects rather than true gaze cues, leading to degraded performance when targets are not salient. To address this, we propose a novel training mechanism to enhance gaze reasoning in VFMs for gaze following. Our method includes: (1) a head-conditioned local LoRA, which enables localized adaptation to preserve scene token learning while improving head token learning for gaze reasoning; and (2) an out-of-cone penalty, which injects gaze cues into head tokens while aligning them with scene tokens. Experiments on the GazeFollow and VAT datasets demonstrate that our method achieves state-of-the-art performance, with particularly strong improvements when gaze targets are not semantically salient. Our findings offer valuable insights for advancing future gaze following research. We will release the code once the paper is accepted.

Object detection from Unmanned Aerial Vehicles (UAVs) is challenged by severe ego-motion, camera jitter, and large scale variations. While modern detectors perform well on static images, their direct application to UAV video often fails, particularly for small objects in dynamic scenes. Existing motion-based methods either rely on computationally expensive optical flow or use single-interval differencing, which is sensitive to jitter and limited in capturing diverse motion patterns. We propose a vision-only motion-guided detection framework that decouples target motion from camera-induced disturbances. A homography-based Global Motion Compensation (GMC) first aligns adjacent frames. We then introduce a Dual-Interval Motion Extraction strategy that captures both short-term and long-term motion cues. To integrate these cues, a lightweight Motion-Guided Attention (MGA) module enhances feature representations within a Feature Pyramid Network. Experiments on the VisDrone-VID dataset demonstrate consistent improvements over a strong YOLOv8 baseline under severe ego-motion. Ablation studies further confirm the effectiveness of the dual-interval design and the proposed motion-guided attention mechanism.

Frozen Vision Foundation Models (VFMs) with lightweight classification heads are increasingly used in medical imaging because they offer efficient and reproducible deployment. Yet noisy-label learning methods for this frozen-feature regime remain poorly understood, and most existing methods still rely on a small-loss assumption inherited from end-to-end training. We present a controlled benchmark of eight noisy-label methods across five medical datasets, three backbones, two noise types, and five noise rates (150 conditions, 6,000 training runs), evaluated with balanced accuracy. The benchmark shows that there is no universal winner: Friedman ranking over the 150 conditions yields $χ^2 = 333.2$ ($p = 4.77 \times 10^{-68}$), ELR wins the most conditions (49/150), while CUFIT attains the best mean rank (2.51). The practical cost of method choice grows sharply with noise severity, from 4.5pp on clean data to 18.8pp at asymmetric 40\% noise. To explain these benchmark-level patterns, we revisit the small-loss assumption in a representative high-risk regime. Under frozen DINOv2 features, clean and noisy loss distributions overlap by 53--61\%, and matched-rate clean-sample detection shows that prediction agreement is markedly more stable than loss ranking under asymmetric noise (3pp vs.\ 13pp precision drop). On ISIC2019 with asymmetric 40\% noise, Co-Teaching reaches 68\% overall accuracy while collapsing to 35.1\% balanced accuracy with zero recall on three minority classes. Together, these results recast noisy-label learning for frozen VFMs as a regime-aware method-selection problem rather than a search for a single dominant algorithm. We conclude with evidence-based guidance and a low-regret feature-space selector for practical recommendation.

Many public buildings provide floorplans with a "you are here" indicator to help visitors orient themselves. Floorplan localization seeks to computationally replicate this capability by determining where visual observations were captured within a floorplan. However, existing methods typically assume controlled small-scale environments and precise vectorized floorplans, limiting their ability to operate in large-scale buildings and rasterized floorplans. In this work, we present an approach for performing floorplan localization in the wild by grounding the task in a reconstructed 3D representation of the scene. Given an unconstrained image collection, our method reconstructs a gravity-aligned 3D scene and projects it into a 2D density map that serves as a floorplan proxy. Floorplan localization is then formulated as aligning this proxy with the input floorplan via a 2D similarity transform. To bridge the appearance gap between density maps and architectural floorplans, we adapt a 2D foundation model to learn cross-modal correspondences, introducing a fine-tuning scheme that encourages semantically aligned matches while preserving structural consistency. Extensive experiments demonstrate substantial improvements over prior methods, including in extremely sparse settings with as little as a single input image. Our code and data will be publicly available.

Online map estimation is a crucial component of autonomous driving systems that reduces the reliance on costly high-definition maps. State-of-the-art (SOTA) methods commonly predict map elements as ordered sequences of points that form polylines and polygons. The evaluation of these methods relies predominantly on mean average precision (mAP) based on thresholded Chamfer distance (CD). This framework lacks sensitivity to point ordering and provides limited granularity in assessing geometric quality, making it difficult to distinguish which methods truly excel over others. In this work, we address these limitations on two fronts. For the single-instance similarity measure, we introduce sequence optimal sub-pattern assignment (SOSPA), an order-aware metric that enables fine-grained evaluation of individual geometries while satisfying all metric axioms. For the multi-instance evaluation framework, we propose polyline localisation and detection (PLD), a soft metric that jointly captures detection quality and geometric accuracy, replacing the hard thresholding of mAP with a principled soft assignment. Through evaluations on nuScenes, we demonstrate that PLD effectively ranks SOTA online mapping methods (MapTRv2, StreamMapNet, MapTracker) while providing a decomposed error analysis. This analysis identifies detection capability as the dominant bottleneck in current methods, revealing a performance trend that mAP fails to capture. Code for evaluation using our metrics will be released.

Accurate segmentation of brain tumour sub-regions from multi-parametric MRI is critical for treatment planning yet remains challenging due to morphological variability, class imbalance, and overlapping appearances of tumour regions across imaging sequences. We propose SegGuidedNet, a three-dimensional residual encoder--decoder network introducing a novel SegAttentionGate module that explicitly supervises the decoder to produce spatially discriminative attention maps for each tumour sub-region necrotic core, peritumoral oedema, and enhancing tumour via a lightweight auxiliary loss, adding less than 0.2% parameter overhead. This sub-region supervision maintains decoder discriminability between visually ambiguous classes while providing free-of-cost spatial interpretability at inference without any post-hoc explanation method. Evaluated independently on BraTS2021 and BraTS2023 GLI across 251 held-out subjects each, SegGuidedNet achieves mean Dice of 0.905 (ET= 0.873, TC=0.906, WT=0.935) and 0.897 (ET=0.859, TC=0.902, WT=0.931) respectively, surpassing ensemble-based nnU-Net and HNF-Netv2 as a single model and approaching Swin UNETR a 10-model ensemble within 2--4 Dice points at a fraction of the inference cost. The consistency of results across two benchmark editions further confirms the generalisability of the proposed approach, offering competitive accuracy with built-in interpretability in a lightweight, clinically practical framework.

Spatio-temporal reasoning is a core capability for Multimodal Large Language Models (MLLMs) operating in the real world. As such, evaluating it precisely has become an essential challenge. However, existing spatio-temporal reasoning benchmark datasets primarily rely on static image sets or passively curated video data, which limits the evaluation of fine-grained reasoning capabilities. In this paper, we introduce VGenST-Bench, a video benchmark that employs generative models to actively synthesize highly controlled and diverse evaluation scenarios. To construct VGenST-Bench, we propose a multi-agent pipeline incorporating a human quality control stage, ensuring the quality of all generated videos and QA pairs. We establish a comprehensive 3x2x2 video taxonomy, encompassing Spatial Scale, Perspective, and Scene Dynamics to span diverse scenarios. Furthermore, we design a hierarchical task suite that decouples low-level visual perception from high-level spatio-temporal reasoning. By shifting the paradigm from passive curation to active synthesis, VGenST-Bench enables fine-grained diagnosis of spatio-temporal understanding in MLLMs.

Training Deep Neural Networks for tracking individual cells in biomedical videos requires a large amount of annotated data. The annotation of videos for cell tracking is very time consuming and often requires domain expertise; this explains the limited availability of public annotated data to address important medical problems like tissue repair or cancer treatment. Generating synthetic videos along with their Ground Truth annotations is a promising solution that relies, as a foundational first step, on the synthesis of single cell annotations (or phantoms). Phantoms need to be time consistent, as they have to replicate biological processes that are specific to the cell types. In this work, we propose a novel framework for generating videos of cell phantoms in the Elliptical Fourier Descriptors (EFDs) domain, a compact and geometrically interpretable representation for 2D closed contours. We represent the cell phantom evolution as a multivariate time series of EFD coefficients, introducing a strong prior for cell morphology and enabling the efficient generation of sequences that evolve coherently in time. Our experimental validation proves that modelling the temporal evolution in EFD space enables the generation of biologically plausible phantom videos. Our method can be used in generative pipelines for synthesizing annotated data for cell tracking, thus strongly mitigating the annotation effort for creating new datasets. Our code is available for download here: https://github.com/FrancescoBenedetto99/efd-cell-video-gen.

Spatio-temporal reasoning in vision-language models requires visual representations that preserve physical geometry rather than merely semantic appearance. Recent multimodal models incorporate geometric information through structural branches, 3D-aware supervision, reasoning-stage fusion, or long-horizon memory. While these approaches demonstrate the importance of geometry for spatial intelligence, they typically treat geometric cues as a shared signal across all visual tokens. We note that this overlooks a finer-grained challenge: different visual tokens require different geometric evidence depending on their spatial roles. To address this limitation, we introduce GeoWeaver, a pre-reasoning geometric grounding framework that treats geometry as a representational prerequisite for spatio-temporal reasoning. GeoWeaver constructs a multi-level geometry bank from a frozen geometry encoder and performs token-adaptive geometric evidence allocation, enabling each visual token to retrieve the most relevant geometric abstractions. The selected evidence is incorporated into visual tokens via a residual grounding operation prior to language modeling, yielding geometry-grounded representations for downstream reasoning. Extensive evaluations on spatial reasoning benchmarks demonstrate that GeoWeaver consistently enhances geometry-aware reasoning while retaining general multimodal capabilities. This indicates that geometric information yields the greatest benefit not as a late-fusion auxiliary signal but as a fundamental prerequisite that shapes the representational foundation on which large language models perform reasoning. All source code and models will be released at https://github.com/yahooo-m/GeoWeaver .

Fashion image retrieval is a cornerstone of modern e-commerce systems. A unified framework that supports diverse query formats and search intentions is highly desired in practice. However, existing approaches focus on narrow retrieval tasks and do not fully capture such diversity. Therefore, in this work, we aim to develop a unified framework capable of handling diverse realistic fashion retrieval scenarios, achieving truly versatile fashion image retrieval. To establish a data foundation, we first introduce U-FIRE, a comprehensive benchmark that consolidates fragmented fashion datasets into a unified collection, supplemented by two manually curated datasets for testing generalization. Building upon this, we propose FashionLens, a unified framework based on Multimodal Large Language Models. To handle divergent matching objectives, we design a Proposal-Guided Spherical Query Calibrator that dynamically shifts query representations into task-aligned metric spaces via adaptive spherical linear interpolation. Additionally, to mitigate the optimization imbalance caused by varying task complexities and data scales, we develop a Gradient-Guided Adaptive Sampling strategy that automatically re-weights tasks based on realtime learning difficulty and the data scale prior. Experiments on U-FIRE show that FashionLens achieves state-of-the-art performance across diverse retrieval scenarios and generalizes robustly to unseen tasks. The data and code are publicly released at https://github.com/haokunwen/FashionLens.

Two-wheelers account for a disproportionately high share of road fatalities in the Global South. Research on two-wheeler rider behavior, however, lags far behind four-wheelers, where multimodal datasets have driven major advances in Advanced Driver Assistance Systems (ADAS). To address this gap, we present the MOtorized TwO-wheeler Rider (MOTOR) dataset, the first large-scale, multi-view, multimodal resource dedicated to two-wheelers in dense, unstructured traffic. MOTOR comprises 1,629 sequences (25+ hours of video data) collected from 16 riders and integrates synchronized front, rear, and helmet videos, rider eye-gaze from wearable trackers, on-road audio, and telemetry (GPS, accelerometer, gyroscope). Rich annotations capture traffic context, rider state, 12 riding maneuvers spanning conventional and unconventional behaviors, and legality labels (Legal, Illegal, Unspecified). We benchmark rider behavior recognition and maneuver legality classification using state-of-the-art video action recognition backbones (CNN and Transformer-based), extended with multimodal fusion, and find that combining RGB, gaze, and telemetry consistently yields the best performance. MOTOR thus provides a unique foundation for advancing safety-critical understanding of two-wheeler riding. It offers the research community a benchmark to develop and evaluate models for behavior analysis, legality-aware prediction, and intelligent transportation systems. Dataset and code is available at https: //varuniiith.github.io/MOTOR-Dataset/

Traditional visual object tracking (VOT) methods typically rely on task-specific supervised training, limiting their generalization to unseen objects and challenging scenarios with distractors, occlusion, and nonlinear motion. Recent vision foundation models, exemplified by SAM 2, learn strong video understanding priors from large-scale pretraining and offer a promising foundation for building more robust and generalizable trackers. However, directly applying SAM 2 to VOT remains suboptimal, as it does not explicitly model target motion dynamics or enforce geometric and semantic consistency across frames, both of which are essential for reliable tracking. To address this issue, we propose SAMOSA, a new tracking framework that adapts SAM 2 to complex VOT scenarios by explicitly leveraging motion, geometry, and semantic cues. Specifically, we introduce a lightweight nonlinear motion predictor to model target dynamics and guide mask selection as well as memory filtering. We further exploit semantic cues to detect target shifts and recover from tracking failures, while geometric cues are incorporated as structural constraints to improve tracking stability. In this way, SAMOSA bridges the gap between the implicit video understanding prior of SAM 2 and explicit tracking-oriented modeling. Extensive experiments show that SAMOSA consistently outperforms state-of-the-art SAM 2--based approaches on general benchmarks, demonstrates stronger generalization than supervised VOT methods, and achieves substantial gains on anti-UAV datasets, which typify complex nonlinear motion scenarios. Our code is available at https://github.com/DurYi/SAMOSA.

Multimodal Large Language Models (MLLMs) have made rapid progress in spatial intelligence, yet existing spatial reasoning benchmarks largely assume pristine visual inputs and overlook the degradations that commonly occur in real-world deployment, such as motion blur, low light, adverse weather, lens distortion, and compression artifacts. This raises a fundamental question: how robust is the spatial intelligence of current MLLMs when visual observations are imperfect? To answer this question, we introduce SpaceDG, the first large-scale dataset for degradation-aware spatial understanding. It is constructed with a physically grounded degradation synthesis engine that embeds degradation formation process into 3D Gaussian Splatting (3DGS) rendering, enabling realistic simulation of nine degradation types. The resulting dataset contains approximately 1M QA pairs from nearly 1,000 indoor scenes. We further introduce SpaceDG-Bench, an human-verified benchmark with 1,102 questions spanning 11 reasoning categories and 9 visual degradation types, yielding over 10K VQA instances. Evaluating 25 open- and closed-source MLLMs reveals that visual degradations consistently and substantially impair spatial reasoning, exposing a critical robustness gap. Finally, we show that finetuning on SpaceDG markedly improves degradation robustness and can even surpass human performance under degraded conditions without any performance drop on clean images, highlighting the promise of degradation-aware training for robust spatial intelligence.

Collaborative driving aims to improve safety and efficiency by enabling connected vehicles to coordinate under partial observability. Recent approaches have evolved from sharing visual features for perception to exchanging language-based reasoning through foundation models for behavioral coordination. Though communicating in language provides intuitive information, it introduces two challenges: high latency caused by autoregressive decoding and information loss caused by compressing rich internal representations into discrete tokens. To address these challenges, we analyze latent communication in collaborative driving under inherent limitations of multi-agent settings. Our analysis reveals agent identity confusion, where direct fusion of latent states entangles decision representations across vehicles. Motivated by this, we propose LACO, a training-free \textbf{LA}tent \textbf{CO}mmunication paradigm that seamlessly adapts pretrained driving models to collaborative settings. LACO introduces Iterative Latent Deliberation (ILD) for latent reasoning, Cross-Horizon Saliency Attribution (CHSA) for communication-efficient information selection, and Structured Semantic Knowledge Distillation (SSKD) to stabilize ego-centric decision making. Closed-loop experiments in CARLA show that LACO notably reduces communication and inference latency while maintaining strong collaborative driving performance.

Fine-grained semantic segmentation requires both precise localization and discrimination between visually similar classes. In FungiTastic, this problem is further complicated by a long-tailed distribution and strong variation in image acquisition conditions. We propose a training-free two-stage framework that decouples segmentation from classification. SAM3 first produces class-agnostic mushroom masks using macro-taxonomic prompts, and DINOv3 then assigns fine-grained labels through prototype matching in the embedding space. To improve this stage, we apply a simple transformation of the DINOv3 feature space that improves prototype-based classification. Compared with class-specific prompting, our approach is more scalable and keeps the segmentation cost low. We report results from one-shot to few-hundred-shot regimes, providing, to the best of our knowledge, the first baseline for fine-grained semantic segmentation in low-data settings.

Vision-Language Models (VLMs) excel at tasks like zero-shot classification and cross-modal retrieval by mapping images and text to a shared space, but this requires expensive end-to-end training with massive paired datasets. Current post-hoc alignment methods reduce computational costs by connecting pretrained encoders through lightweight mappings, yet still demand substantial paired data. In this work, we investigate the potential of repurposing the classification heads of pretrained vision models as semantic prototypes. The recycling of these weights, typically discarded after pretraining, unlocks two distinct capabilities: it enables zero-shot alignment by using weights as semantic anchors, and serves as a robust data augmentation strategy by mixing these prototypes with real image-text pairs. We demonstrate that integrating our approach with several state-of-the-art post-hoc alignment techniques consistently boosts accuracy in cross-modal retrieval, zero- and few-shot classification tasks.

Evaluating single-concept personalization in text-to-image diffusion requires measuring both concept preservation, which captures identity fidelity to a reference, and prompt following, which captures whether the generated scene matches the prompt. Existing metrics commonly compute these signals using global image or text-image embeddings, such as CLIP-I, DINO, and CLIP-T. We show that such metrics correlate poorly with human perception because they attend to the image as a whole instead of separating the concept subject from the background. We introduce MaSC, a masked similarity metric that uses externally provided foreground concept masks to decompose evaluation into subject-specific concept preservation and background-based prompt following. MaSC computes both scores from frozen SigLIP2 SO400M-NaFlex features: concept preservation is measured by masked max-cosine matching between foreground reference patches and generated-image patches, while prompt following is measured by comparing a background-only pooled image embedding to a subject-stripped prompt embedding. On DreamBench++ human ratings, MaSC achieves Krippendorff alpha = 0.471 for concept preservation, outperforming all tested non-LLM baselines and GPT-4V, and approaching GPT-4o. On ORIDa, a real-photo identity-preservation benchmark across physical environments, MaSC achieves AUC = 0.992, nearly perfectly distinguishing same-subject from cross-subject pairs. Its prompt-following score also outperforms the CLIP-T baseline shipped with DreamBench++. These results show that spatially decomposed aggregation is a strong design principle for evaluating concept-driven generation.

We propose SADGE, a quantitative similarity metric that predicts the performance of synthetic image datasets for common computer vision tasks without downstream model training. Estimating whether a synthetic dataset will lead to a model that performs well on real-world data remains a bottleneck in model development. Existing evaluation metrics (e.g., PSNR, FID, CLIP) primarily measure semantic alignment between real and synthetic images (Appearance Similarity Score). Less commonly, structural similarity between images is considered to assess the domain gap (Geometric Similarity Score). However, to the best of our knowledge there exists no studies that evaluate which similarity metric is the best downstream predictor for a given synthetic dataset. In this paper, we show over a wide variety of different synthetic datasets and downstream tasks that neither appearance nor geometry alone can reliably predict downstream performance; rather, it is their non-linear interplay that dictates synthetic data utility. Specifically, we measure how commonly used Appearance and Geometric Similarity metrics computed between synthetic and real images correlate with downstream performance in object detection, semantic segmentation, and pose estimation. Across five public synthetic-to-real benchmark families and 15 dataset-level variants (79k image pairs), SADGE achieves the strongest association with downstream transfer performance under both linear and rank-based criteria, reaching Pearson r=0.88 and Spearman rho=0.77. We compute for each combination of geometry-based methods and appearance-based approaches SADGE scores across all benchmark families. The best configuration is obtained by fusing DINOv3 appearance similarity with MASt3R geometric consistency through a constrained bilinear interaction, outperforming both the strongest geometry-only baseline and the strongest appearance-only baseline .

Real-world deployment of AI vision models is both fueled and limited by the data available for training and testing. Real datasets are sparse and uneven: long-tailed or unbalanced distributions hinder generalization, and the low number of samples in low density regions makes it hard to run evaluations. Synthetic data can fill these gaps, providing us with a way to sample the input space more continuously and improve data coverage for benchmarks. Focusing on the autonomous driving safety-critical case of pedestrian detection in the dark, we show how synthetic low-light samples can be used to better characterize the performance of a state-of-the-art object detection model as a function of the scene illumination. We use a synthetic RAW image augmentation technique to generate low-light samples that match the noise model of the camera sensor. Performance metrics on real and synthetic low-light data are similar, indicating that the AI model finds it hard to distinguish between them.

While large vision-language-action (VLA) models and generative world models (WM) have advanced long-horizon embodied intelligence, their practical deployment remains challenged by uncertainty in learning-based action generation. Low-quality actions may cause physical failures during execution or lead to misleading world-model rollouts with redundant rendering costs. To address this issue, we propose Pre-VLA, a unified runtime verification architecture that performs preemptive action validity assessment before physical execution or world-model imagination. Pre-VLA leverages an efficient multimodal backbone with modality-aware pooling and a lightweight dual-branch head to predict both safety confidence and critic-derived advantage scores for candidate action chunks. To handle severe class imbalance and unstable boundary decisions, we train Pre-VLA with a multi-task objective combining Focal classification, advantage regression, and soft-threshold calibration. During deployment, a dual-mode preemptive resampling scheduler filters low-quality actions and triggers adaptive resampling under a limited computation budget. Experiments on the LIBERO benchmark show that Pre-VLA improves the average closed-loop success rate across four suites from 30.79\% to 37.62\% over RynnVLA-002, reduces task execution steps, achieves 183.9 ms average forward verification time per action chunk, and mitigates error accumulation in world-model rollouts.

Remote photoplethysmography (rPPG) enables non-contact measurement of cardiac pulse signals by analyzing subtle color changes in facial videos. Nevertheless, extracting rPPG signals remains challenging because of their extremely weak signal strength and susceptibility to illumination noise. In this paper, we propose an rPPG signal extraction method that exploits the quasi-periodic characteristics of rPPG signals. Our approach models quasi-periodicity of the rPPG signal, which arises from the stable cardiac cycle, as a block-sparse structure in the time-frequency domain. To incorporate a block-sparse model and enable adaptive signal separation under illumination fluctuations, we construct a time-varying signal separation framework. Experiments using a public dataset demonstrate the effectiveness of our method.

Table structure recognition (TSR) requires both table-level coherence (row/column counts, headers, spanning cells) and precise separator localization. We introduce FastTab, a grid-centric TSR model that avoids autoregressive HTML decoding by combining (i) a lightweight Tiny Recursive Module (TRM) for global reasoning and (ii) axial 1D Transformer encoders that capture long-range dependencies along rows and columns. The model predicts row/column counts, header rows, and separators to construct a grid, then infers rowspan/colspan using ROI-aligned cell features. Across four benchmarks (PubTabNet, FinTabNet, PubTables-1M, and SciTSR), FastTab achieves competitive structure recovery performance while operating at low-latency inference. We further study robustness under pixel-level anonymisation and show an extension to curved separators for camera-captured documents. The source code will be made publicly available at https://github.com/hamdilaziz/FastTab .

Urban scene reconstruction from real-world observations has emerged as a powerful tool for self-driving development and testing. While current neural rendering approaches achieve high-fidelity rendering along the recorded trajectories, their quality degrades significantly under large viewpoint shifts, limiting the applicability for closed-loop simulation. Recent works have shown promising results in using diffusion models to enhance quality at these challenging viewpoints and distill improvements back into 3D representations. However, they often require costly per-scene optimization, and the distilled representations remain fragile and fail to generalize beyond limited synthesized views. To address these limitations, we propose GenRe, a novel diffusion-guided generalizable enhancer for urban scene reconstruction. GenRe takes as input any pretrained 3D Gaussian representation and fixes the deficiencies within a few minutes. By learning to distill generative priors across diverse scenes, GenRe produces robust and high-fidelity representation efficiently that generalizes reliably to challenging unseen viewpoints (e.g., lane change). Experiments show that GenRe outperforms existing methods in both quality and efficiency and benefits various downstream tasks, enabling robust and scalable sensor simulation for autonomous driving.

Visual Question Answering (VQA) holds great promise for clinical support, particularly in ophthalmology, where retinal fundus photography is essential for diagnosis. However, ophthalmic VQA benchmarks primarily emphasize answer accuracy, neglecting the explicit visual evidence necessary for clinical interpretability. In this work, we introduce FundusGround, a new benchmark for clinically interpretable ophthalmic VQA with spatially-grounded lesion evidence. Specifically, we propose a three-stage pipeline that collects 10,719 fundus images with 15,595 image-level meticulously annotated lesions. To ensure anatomical consistency and clinical validity, all lesions are spatially localized using the Early Treatment Diabetic Retinopathy Study (ETDRS) grid, enabling standardized mapping to nine clinically meaningful retinal regions. Built upon this structured lesion evidence, 72,706 questions are then generated spanning four formats: open-ended, closed-ended, single-choice, and multiple-choice. We further benchmark multiple general- and medical- large vision-language models using dual metrics for answer accuracy and lesion-level reasoning. The experiments demonstrate that incorporating lesion-level visual evidence consistently improves model performance and transparency, highlighting the necessity of explicit spatial grounding for reliable and explainable ophthalmic VQA.

Extracting structured information from visual documents (Visual Information Extraction, VIE) is a cornerstone of business automation. While recent Multimodal Large Language Models (MLLMs) have shown promising capabilities, existing benchmarks suffer from critical limitations in scale and realism, lack semantic granularity, and fail to cover diverse document types. To bridge this gap, we introduce ReceiptBench, a large-scale, human-annotated benchmark consisting of 10k diverse receipts, organizing information extraction into four hierarchical sub-tasks: (1) Basic Perception for raw text spotting, (2) Format Normalization for strictly following standardization instructions, (3) Semantic Reasoning for inferring implicit attributes from context, and (4) Structure Parsing for handling nested line items. Furthermore, we propose a two-stage training framework incorporating Metric-Aware Group Relative Policy Optimization (GRPO), which translates rigorous evaluation constraints into reinforcement learning signals to enhance structural consistency. Extensive experiments demonstrate that our method yields state-of-the-art performance, surpassing leading proprietary models on complex reasoning tasks. We release our datasets and code at https://github.com/wwwT0ri/ReceiptBench.

Surface electromyography (sEMG) signal-based activity recognition has attracted increasing research attention in recent years. To develop accurate sEMG signal-based activity recognizers, numerous approaches have been proposed. Some studies focus on designing larger and more expressive model architectures to enhance the representational capacity of sEMG signals, while others aim to enrich model priors through large-scale pretraining, thereby improving recognition performance. Recently, large language models (LLMs) have shown remarkable generalization and reasoning capabilities in natural language processing, whose implicit knowledge, learned from extensive linguistic descriptions of actions, opens new possibilities for interpreting sEMG signals and inferring activity intentions. Motivated by this, we propose LLM-sEMG, a novel framework that leverages LLMs as sEMG activity recognizers. Within this framework, we design a language-oriented mapping mechanism that converts continuous sEMG sequences into sEMG language, integrating several strategies to further facilitate the signal-to-language mapping process. Extensive experiments demonstrate that the proposed framework achieves highly accurate sEMG signal-based activity recognition using large language models.

Agricultural decision-making increasingly requires multimodal systems that can transform visual observations into reliable, executable actions. However, existing agricultural multimodal benchmarks mainly evaluate final-answer correctness and provide limited support for assessing whether models can use external tools to complete precision-sensitive workflows. In this paper, we introduce AgroTools, a benchmark for evaluating tool-augmented multimodal agents in agriculture. AgroTools contains 539 question-answer instances paired with 1,097 heterogeneous agricultural images, spanning five task families and an executable environment of 14 agricultural tools. Each query is annotated with structured tool-use traces, enabling a dual-view evaluation of both process-level execution quality and outcome-level task success. We benchmark 9 open-source and 4 closed-source multimodal large language models on AgroTools. Results show that current models remain far from reliable in agricultural tool-use settings, with clear bottlenecks in tool planning, argument generation, execution recovery, and final-answer synthesis. We hope AgroTools will support future research on multimodal agents for high-precision agricultural applications. The benchmark and evaluation are available at https://huggingface.co/datasets/AgroTools/AgroTools.

Eye tracking (ET) is a foundational technology for advanced AR/VR applications. However, training ET models for every new ET device is challenging: real data collection is costly and time-consuming, while existing synthetic data generation methods lack realism. To remove the need for additional data collection while maintaining data quality, we introduce a data-driven 3D prior that models the distribution of human eyes across diverse identities, gaze directions, and light settings. This model, which we coin GazePrior, then enables sparse-input 3D reconstruction of annotated data collected with previous ET devices, which can in turn be rendered from the cameras of any target ET device. Our approach synthesizes data with the realism, diversity and ground-truth accuracy of real data collection without its prohibitive costs. Our experiments demonstrate that ET models trained with our synthesized data outperform previous zero-shot methods, achieving higher accuracy and robustness.

Accurate segmentation of hepatic and portal vessels in contrast-enhanced computed tomography angiography (CTA) remains challenging due to complex vascular topology, peripheral visibility limitations, and acquisition-induced ambiguities. While existing public datasets offer valuable benchmarks, few include clinically realistic annotation constraints. We introduce VEELA (Vessel Extraction and Extrication for Liver Analysis), a rigorously curated liver vessel dataset derived from 40 CTA scans inherited from the CHAOS grand-challenge cohort. All vessels were manually delineated slice-by-slice under multi-expert consensus, using a strict visibility-driven annotation policy and avoiding anatomically inferred interpolation. This design explicitly captures anatomical variability and imaging-related uncertainty. As a continuation of the CHAOS challenge, VEELA enables reproducible cross-benchmark evaluation while extending the scope to fine-grained hepatic and portal vessel segmentation. We further establish a standardized benchmarking framework and analyze complementary evaluation metrics, including topology-aware (clDice), overlap-based (IoU), boundary-sensitive (NSD), and geometry-aware (area, length) measures. Our results demonstrate that different metrics capture distinct aspects of vascular integrity, underscoring the necessity of multi-perspective evaluation for clinically meaningful vessel segmentation. VEELA is publicly released to facilitate reproducible research and support the development of robust vascular segmentation methods. Researchers can access the evaluation metrics, dataset, and submission platform at https://www.synapse.org/Synapse:syn65471967.

Low-bit quantization is widely used to compress super-resolution (SR) models and reduce storage and computation costs for deployment on resource-limited devices. However, when SR models are pushed to ultra-low precision (2-4 bits), performance can drop sharply due to diminished representational capacity and the detail-sensitive nature of SR. To address these issues, we propose QuantSR+, a unified framework that improves quantization operators, network design, and training optimization, achieving better trade-offs between accuracy and efficiency than prior low-bit SR methods. QuantSR+ mainly relies on three technical contributions: (1) Redistribution-driven Bit Determination (RBD), which reshapes quantization distributions in both forward and backward passes to preserve representation fidelity; (2) Quantized Slimmable Architecture (QSA), which begins with an over-parameterized model and progressively prunes less critical blocks to meet efficiency budgets while pushing the accuracy performance; and (3) Slimming-guided Function-localized Distillation (SFD), which enforces block-aware feature alignment via a direct loss and a progressive, function-local training schedule to capture quantization effects better and speed up convergence. Extensive experiments show that QuantSR+ achieves state-of-the-art performance against both specialized quantized SR methods and generic quantization approaches. For SwinIR-S on Urban100 (x4), it improves PSNR by 0.29 dB over the 2-bit SOTA baseline. Meanwhile, it delivers strong efficiency gains at 2-bit, reducing operations by up to 87.9% and storage by 89.4%. QuantSR+ is effective for both convolutional and transformer-based SR models, indicating broad applicability.

Multimodal large language models (MLLMs) and diffusion models have each reached remarkable maturity: MLLMs excel at reasoning over heterogeneous multimodal inputs with strong semantic grounding, while diffusion models synthesize images and videos with photorealistic fidelity. We argue that these two families can be unified through a simple division of labor: MLLMs perform semantic planning, while diffusion models render pixels from high-level semantic guidance and low-level visual features. Building on this idea, we propose Bernini, a unified framework for video generation and editing. An MLLM-based planner predicts the target semantic representation directly in the ViT embedding space, and a DiT-based renderer synthesizes pixels conditioned on this plan, augmented by text features and, for editing, source VAE features for detail preservation. Because semantics serve as the interface, the planner and renderer can be trained separately and only lightly co-trained, preserving the pretrained strengths of both components while keeping training efficient. To better handle multiple visual inputs, we introduce Segment-Aware 3D Rotary Positional Embedding (SA-3D RoPE), and further incorporate chain-of-thought reasoning in the planner to better transfer understanding into generation. Bernini achieves state-of-the-art performance across a wide range of video generation and editing benchmarks, with the MLLM's pretrained understanding translating into strong generalization on challenging editing tasks.

While 4D Gaussian Splatting (4DGS) has revolutionized high-fidelity dynamic reconstruction, safeguarding the intellectual property of these assets remains an open challenge. Conventional steganographic techniques often neglect the underlying kinematic manifolds, triggering non-physical artifacts such as severe temporal flickering and "FVD collapse". To address this, we propose \textbf{4D-GSW}, a kinematic-aware watermarking framework designed to embed robust copyright information while preserving high spatio-temporal consistency. Unlike prior 4D steganography that primarily focuses on opacity-guided invisibility, our approach explicitly addresses the physical coherence of motion trajectories. We introduce a \textbf{Spatio-Temporal Curvature (STC)} metric to identify "Dynamic Instants," adaptively gating watermark gradient injection to shield critical motion manifolds from non-physical perturbations. To ensure global coherence across complex deformations, we formulate a joint \textbf{HMM-MRF energy minimization} model that synchronizes watermark phases within both temporal trajectories and spatial neighborhoods. Furthermore, an \textbf{anisotropic gradient routing} mechanism ensures that watermark embedding remains strictly decoupled from photometric reconstruction fidelity. Extensive experiments have demonstrated the superior performance of our method in robustly hiding watermarks while resisting various attacks and maintaining high rendering quality and spatiotemporal consistency.

Land Use Land Cover (LULC) classification is essential for national 3D mapping, geospatial analysis, and sustainable planning. Multispectral (MS) LiDAR provides synchronized spatial-spectral information, and deep learning (DL) enables 3D point cloud semantic segmentation; however, adoption is limited by the lack of publicly available urban and suburban MS LiDAR datasets aligned with National Mapping and Cadastral Agencies (NMCAs) classification schemes. This study addresses these gaps by introducing L1 and L2 NMCA-aligned LULC classification schemes and a new benchmark MS LiDAR dataset. We evaluate seven state-of-the-art DL models and perform spectral ablation studies at both levels of detail. Results show that Point Transformer V3 achieves the best performance, with mIoU of 79.4% (L1, 8 classes) and 58.9% (L2, 20 classes) using a dual-wavelength LiDAR system (532 nm and 1064 nm). Ablation results show that multispectral information improves performance over geometry-only inputs, with gains of 1.1 percentage points at L1 and 7.8 points at L2. These results highlight the value of LiDAR reflectance for fine-grained material discrimination and support the evolution of NMCA LULC schemes toward higher semantic detail. The Loosdorf-MSL dataset contributes a new benchmark for consistent national and international LULC mapping.

Purpose: To assess whether breast lesion segmentation can be learned directly from acquired MRI k-space, and whether doing so improves robustness when data are accelerated or noisy. Materials and Methods: This retrospective study used public breast dynamic contrast-enhanced MRI (DCE-MRI) datasets with acquired and synthetic k-space, together with a within-dataset synthetic control. We compared four 3D U-Net variants: a hybrid k-space-to-image model, a native k-space model, and magnitude and complex image-space baselines. Models were evaluated under increasing undersampling and added complex Gaussian k-space noise. The primary outcome was patient-level Dice similarity coefficient under cross-validation, with the hybrid model prespecified as the main comparison against the magnitude image-space baseline. Results: At full sampling, the hybrid and image-space models performed similarly. As acceleration increased, the hybrid model retained substantially more segmentation accuracy and significantly outperformed the magnitude image-space baseline across moderate to high undersampling levels. The same pattern was observed when noise was added directly to k-space: the hybrid model degraded more slowly, whereas the image-space baseline failed under heavier noise. This advantage was reproduced in the within-dataset synthetic control. Feature analysis suggested that the k-space stage and image-space stage played complementary roles, with frequency-domain filtering concentrated before image-domain lesion localization. Conclusion: K-space-aware deep learning improves the robustness of breast lesion segmentation under MRI undersampling and k-space noise, while matching image-space methods at full sampling.

Identity-conditioned diffusion models enable high-quality and identity-consistent face generation, but they also raise severe privacy concerns, as models may continue to synthesize individuals despite their right to be forgotten. While machine unlearning has been extensively studied for concept and data removal, identity unlearning remains largely unexplored, particularly in models conditioned directly on identity embeddings rather than text prompts. In this work, we study identity unlearning in Arc2Face, a state-of-the-art identity-conditioned latent diffusion model for face generation, and introduce Proximity-guided Identity Unlearning (PIU), an anchor-guided framework for identity unlearning. Specifically, we formulate identity removal as an identity replacement objective that reassigns the source identity to a selected anchor identity in the learned identity space, and we complement it with a proximity-based anchor selection strategy motivated by the geometry of ArcFace representations. We further show that effective unlearning can be achieved through localized fine-tuning of a small subset of identity-sensitive cross-attention layers. Experiments across many target identities show that our framework effectively suppresses generation of the target identity while preserving realism and identity consistency for retained identities, as validated by improved performance on unlearning and image-quality metrics, together with qualitative evaluation. The source code for the PIU framework is publicly available at https://github.com/edgarcancinoe/piu_unlearning .

Accurate detection and counting of virus patches in focus-forming unit (FFU) images, also known as foci images, are important for quantifying viral infection and analyzing cellular structures. This task is challenging because biomedical targets often vary substantially in size, density, contrast, and shape. In this paper, we propose an enhanced YOLOv2-based detector that integrates a Feature Pyramid Network (FPN) to improve multi-scale feature representation. We also incorporate a switchable atrous convolution mechanism to adapt the receptive field for fine-grained targets in dense microscopy images. The proposed method is evaluated on biomedical foci image datasets for virus patch and small cell patch detection. For small cell patch detection, the model achieves a mean average precision (mAP) of 40.5% at a 25% Intersection over Union (IoU) threshold. For FFU virus patch detection, the model achieves an mAP of 68%. These results indicate that combining FPN-based feature fusion with switchable convolution improves the suitability of YOLOv2 for specialized biomedical object detection tasks

Vision-language models (VLMs) have achieved strong performance across diverse multimodal tasks, but their adversarial robustness in visible-infrared (VIS-IR) scenarios remains underexplored. This gap is critical because VIS-IR sensing is widely used in real-world perception systems to support reliable understanding under challenging imaging conditions. To address this cross-modal threat setting, we propose CFGPatch, a curved-edge fractal geometric adversarial patch framework for attacking VIS-IR VLMs. CFGPatch builds on triangular fractal geometry and replaces rigid straight-edged primitives with Bezier-curved elements, preserving multi-scale fractal self-similarity while introducing smoother contours, richer directional variation, and more flexible shape deformation. In addition, we design a modality-specific Fraser-spiral rendering mechanism to inject fine-grained texture distortions and misleading perceptual cues into visible and infrared images. By coupling global curved-fractal geometry with local spiral-based appearance interference, CFGPatch disrupts both shape perception and texture interpretation. We further adopt expectation over transformation (EOT) to improve robustness against common image-level transformations. Extensive experiments show that CFGPatch effectively fools VIS-IR VLMs and consistently outperforms standard patch baselines in attack effectiveness and robustness. Moreover, adversarial samples optimized for zero-shot classification transfer well to image captioning and visual question answering, demonstrating strong cross-task transferability and generalizability across downstream tasks.

Long streaming video QA remains challenging due to growing visual tokens and limited reasoning length of large language models (LLMs). KV-caching stores the Key-Value (KV) of the historical tokens via LLM prefill and enables more efficient streaming QA. However, existing methods cache every one or two frames, causing redundant memory usage and losing fine-grained spatial details within frame or temporal contexts across frames. This paper proposes MuKV, a method that features a multi-grained KV cache compression module and a semi-hierarchical retrieval approach to improve both efficiency and accuracy for long streaming VideoQA. For the offline KV cache, MuKV extracts visual representations at patch-, frame-, and segment-levels. The multiple levels of granularity preserve both local cues and global temporal context, while maintaining efficiency with a dual signal token compression mechanism guided by self-attention and frequency. For online QA, MuKV designs a semi-hierarchical retrieval method to retrieve relevant KV caches for answer generation. Experiments on long-streaming VideoQA benchmarks show that MuKV significantly improves answer accuracy, without sacrificing memory and online QA efficiency. Moreover, our compression mechanism alone brings consistent benefits across answer accuracy, memory, and QA efficiency over baselines, showcasing highly effective contribution.

Earth Observation (EO) imagery is often degraded by atmospheric turbulence and pointing jitter; yet, these effects are rarely considered in datasets used to train AI-based detection models. Based on prior work, this paper presents an enhanced image simulator that enables the incorporation of vertical-path atmospheric turbulence and satellite pointing jitter, arising from platform and sensor vibrations, to generate physically realistic distorted images. As a case study, vessel detection is evaluated using YOLOv8 and RetinaNet on images generated by the proposed simulator under different levels of turbulence and pointing errors. Results show that YOLOv8 recall decreases from 91% under ideal conditions to 60% in the presence of weak turbulence, and falls below 40% under strong turbulence or jitter. In contrast, RetinaNet demonstrates greater robustness, maintaining approximately 75% recall across degraded conditions. These results highlight the importance of incorporating realistic physical degradations into EO training datasets to ensure reliable performance of AI-based models in operational environments, as demonstrated in maritime surveillance applications.

Heterogeneous sensor fusion is vital for detecting, localizing, and classifying CBRNE threats. However, individual sensors are often only capable of detecting a subset of relevant threats with varying reliability or can even provide only indirect threat indications, making threat classification challenging. Furthermore, high clutter rates on the sensor side present a great challenge for fusion systems. Additionally, the limited availability of high quality datasets hinders the advancement of learning-based detection and classification models in smart sensors. To mitigate these sensor related shortcomings, a context-aware and domain knowledge-enhanced fusion process is proposed. First, a novel evidence hierarchy is established that enables modeling of direct, indicative, and contextual information. Second, contextual information about the environment is introduced into the fusion process, by collecting, processing, and exploiting OSINT inputs. Third, all levels of the evidence hierarchy are used to craft a Bayesian threat type classification mechanism with domain knowledge-informed priors. The proposed methodology is evaluated in simulated scenarios, and the results demonstrate the benefit of the proposed fusion approach in terms of robustness to clutter and prior mismatch, with an overall classification accuracy of up to 95%.

Accurate brain tumor segmentation using multiparametric MRI is critical for effective treatment planning. However, in clinical settings, complete acquisition of all MRI sequences is not always possible. The absence of certain MRI modalities results in substantial performance degradation in existing segmentation methods, which typically rely on naive feature concatenation or direct fusion strategies. To address this limitation, we propose a novel segmentation model D3Seg which is designed to maintain stable performance under missing-modality settings. D3Seg introduces Multi-hop Modality Graph Fusion (MMGF) to model higher order inter-modality dependencies, a lightweight diffusion-based imputation mechanism to compensate for missing T1ce representations in latent space, and probability-space decision refinement to mitigate dominant class overconfidence and improve delineation of underrepresented tumor subregions. Extensive evaluation on BraTS 2023 dataset demonstrates that our D3Seg model consistently improves segmentation performance under missing modality configurations. The proposed model achieves approximately 1.5-2.0% Dice improvement on enhancing tumor (ET) and around 1.0% on tumor core (TC) across multiple missing modality configurations compared to the current state-of-the-art model, while maintaining computational efficiency.

We introduce a novel 3D hand pose estimator that can accurately recover the shape and pose of people's hands in a room from afar, typically from fixed cameras at room corners, in extremely low-resolution and frequently occluded views. Our key idea is to fully leverage hand-body coordination, its temporal progression, and multiview observations. We achieve this with a novel Transformer-based model, in which hand and body configurations are modeled through correlations between their visual features expressed as per-view tokens, and their temporal coordination is exploited in an autoregressive manner. We introduce a novel dataset, which we refer to as REACH, Room-Environment dataset Annotated with Chest cameras for Hand pose estimation, to train and test our method. REACH is a first-of-its-kind large-scale hand pose dataset that captures accurate hand movements of 50 participants across a wide variety of daily activities. In order to avoid interfering with natural movements while annotating the hands with accurate shape and pose, we leverage concealed chest cameras. Through extensive experiments, including comparative studies with existing methods, we show that our model, REACH-Net, achieves highly accurate 3D hand pose estimation from afar. These results broaden the horizon of 3D hand pose estimation, especially towards "in-the-wild" continuous human behavior analysis.

Video Capsule Endoscopy (VCE) poses a challenging multi-label temporal classification problem, requiring simultaneous localization of 8 anatomical regions and detection of 9 pathological findings across tens of thousands of frames. We present GALAR-TemporalNet v2, a hierarchical temporal model that addresses three core challenges: extreme class imbalance, long-range temporal dependencies, and pathology--anatomy entanglement. Our architecture combines windowed self-attention for local modeling, a Dual-Graph GCN for global frame relationships, and Bidirectional Mamba for selective boundary context encoding. A novel anatomy prototype residual pathway decouples pathological deviation signals from normal organ appearance, and a frame-level GCN skip connection stabilizes training of visually confusable rare classes. The competition version, GALAR-TemporalNet, achieved an overall mAP@0.5 of 0.2644 and mAP@0.95 of 0.2353 on the RARE-VISION test set. Following the competition, the redesigned GALAR-TemporalNet v2 -- incorporating a restructured pathology branch, refined loss functions, and extended post-processing -- improved these results to mAP@0.5 of 0.3409 and mAP@0.95 of 0.3333.

Zero-shot temporal action localization (ZS-TAL) consists of classifying and localizing actions in untrimmed videos, where action classes are unseen at training time. Existing work uses Vision and Language Models (VLMs), taking advantage of their strong zero-shot transfer capabilities. Yet, these models face evident challenges with fine-grained action classification, making it difficult to directly use them to distinguish between the presence and absence of an action. Most current methods for ZS-TAL address these challenges by training models on large-scale video datasets, which require annotated data and often result in limited generalization performance. Recently, approaches discarding the use of labeled data have emerged as an alternative. Following this direction, we propose a novel approach, ``Textual Guidance for finer localization of actions in videos'' (TEGU), that compensates for the lack of supervision from training data by exploiting rich textual information derived from large language models and structured text extracted from captions. This additional linguistic context can improve fine-grained discrimination by providing richer cues about fine-grained action differences within videos. We validate the effectiveness of the proposed method by conducting experiments on the THUMOS14 and the ActivityNet-v1.3 datasets. Our results show that, by exploiting rich textual information for improved action localization, TEGU outperforms state-of-the-art ZS-TAL approaches that do not involve training

Achieving high levels of surgical skill through effective training is essential for optimal patient outcomes. Automated, data-driven skill assessment holds significant potential to improve surgical training. While machine learning-based methods are increasingly popular for assessing skills in minimally invasive surgery, their application to open surgery remains limited. We present the results of a dedicated MICCAI challenge designed to benchmark and advance vision-based skill assessment in open surgery. The challenge dataset comprises videos of an open suturing training task recorded with a static GoPro camera in a dry-lab setting, with instrument trajectories available in addition to the primary video modality. The OSS Challenge was hosted over two consecutive years, comprising two and three independent tasks, respectively: (1) classifying skill level into four classes, (2) predicting the full Objective Structured Assessment of Technical Skills across eight categories, and (3) tracking hands and surgical tools. Participants submitted diverse solutions including deep learning-based video models, tracking-driven methods, and hybrid approaches. General-purpose spatiotemporal video models consistently achieved the strongest performance, though conceptually diverse approaches reached competitive levels when well-executed. Predicting fine-grained OSATS scores remains challenging but benefits substantially from increased training data. Keypoint tracking proves difficult given frequent occlusions and out-of-frame instances, limiting current applicability for motion-based skill analysis. This work benchmarks innovative and diverse solutions for surgical skill assessment, highlighting both the promise and current limitations of video-based evaluation in open surgery and identifying critical directions for advancing automated skill assessment toward clinical impact.

Blind image quality assessment (BIQA) for ultrahighdefinition (UHD) images remains challenging because native-resolution inference is computationally expensive, whereas aggressive resizing or isolated cropping may suppress scale-sensitive distortions and weaken the relationship between local artifacts and global scene context. This paper aims to improve UHD-BIQA by explicitly modeling the structural dependencies among sampled image regions rather than treating them as independent views, and a graph representation learning framework UHD-GCN-BIQA is proposed. The framework samples aspect-ratio-aligned patches from each UHD image, encodes them as graph nodes, and constructs a hybrid k-nearest-neighbor graph using spatial proximity and feature similarity. Residual graph convolution is used to propagate contextual information across regions, and gated attention pooling aggregates patchlevel evidence into an imagelevel quality prediction. An exponential moving average normalized multiobjective loss function is adopted to stabilize the joint optimization of regression, correlation, and ranking objectives. Experiments on the UHD-IQA benchmark show that UHD-GCN-BIQA achieves PLCC = 0.7784, SRCC = 0.8019, and RMSE = 0.0519, obtaining competitive correlation performance and the lowest RMSE among the compared methods. These results indicate that graph-based region relation modeling is effective for UHD image quality assessment, particularly for improving absolute quality score estimation under high-resolution visual content.

Recent feed-forward 3D gaussian splatting methods have made dramatic progress on individual aspects of 3D scene reconstruction, but no existing method jointly addresses dynamic content, multi-view input, and unknown camera poses in a single feed-forward pass. Methods that handle dynamics either require accurate camera poses or accept only monocular input; pose-free multi-view methods address only static scenes; and per-scene optimization methods bridge some of these gaps but at minutes-to-hours cost per scene. We introduce NoPo4D, the first feed-forward system that addresses this empty quadrant. Building on a pretrained geometry backbone and recent 4D Gaussian frameworks, NoPo4D introduces a velocity decomposition that splits Gaussian motion into per-pixel image-plane shifts and depth changes, allowing direct supervision from pseudo ground-truth optical flow on the 2D component. This sidesteps both the differentiable rendering that couples prior posed methods to pose accuracy and the 3D motion ground truth that prior pose-free methods require. The system is rounded out by a bidirectional motion encoder for cross-view and cross-frame feature aggregation, and view-dependent opacity that mitigates cross-view and cross-timestep Gaussian misalignments. On four multi-view dynamic benchmarks, NoPo4D consistently outperforms prior feed-forward baselines, and with an optional post-optimization stage surpasses per-scene optimization methods, while running orders of magnitude faster.

Event-based low-light image enhancement (LIE) methods mainly focus on incorporating high dynamic range (HDR) information from events while overlooking the essential global illumination in images and the inherent noise sensitivity of event signals in real-world scenarios. To address these issues, we propose EIC-LIE, an event-illumination collaborative LIE framework. Concretely, we first design an Event-Illumination Collaborative Interaction (EICI) module, which contains two key processes: forward gathering, which gathers HDR features across varying lighting conditions, and backward injection, which provides complementary content for illumination and event representations. Next, we introduce an Illumination-aware Event Filter (IAEF) that dynamically reduces event noise based on brightness statistics derived from images. Additionally, we build a beam-splitter-based hybrid imaging system to collect high-quality event-image pairs with temporal synchronization from dynamic scenes, providing the first high-resolution, real-world event-based LIE dataset. Extensive experiments show that our EIC-LIE outperforms state-of-the-art methods on five real-world and synthetic datasets, significantly surpassing previous methods with improvements of up to 1.24dB in PSNR and 0.069 in SSIM. The code and dataset are released at https://github.com/QUEAHREN/EIC-LIE.

Recent advancements in Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in general visual understanding. However, their application to safety-critical driving scenarios remains limited by an inability to accurately perceive and reason about rare high-stakes dynamic events, such as collisions or near-collisions. To address this, we introduce a pipeline that enhances MLLM perception by fusing downsampled video frames with synchronized high-frequency telematics data (IMU and GPS) and semantic insights from specialized computer vision models. Our pipeline generates high-quality pseudo-labels, including descriptive captions and question-answer pairs, specifically designed to train MLLMs to identify and describe Safety-Critical Events (SCEs) in real-world driving footage. We show the effectiveness of our approach fine-tuning the open-source QwenVL-2.5 model via DoRA adapters: our experiments demonstrate significant improvements in identifying and explaining safety-critical events, with fewer than 50M trainable parameters and limited computational budget.

Deep learning models, including Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), have achieved state-of-the-art performance on various computer vision tasks such as object classification, detection, segmentation, generation, and many more. However, these models are data-hungry as they require more training data to learn millions or billions of parameters. Especially for supervised learning tasks, curating a large number of labeled samples for model training is an expensive and time-consuming task. Active Learning (AL) has been used to address this problem for many years. Existing active learning methods aim at choosing the samples for annotation from a pool of unlabeled samples that are either diverse or uncertain. Choosing such samples may hinder the model's performance as we pool based on one dimension, i.e., either diverse or uncertain. In this paper, we propose four novel hybrid sampling methods for pooling both easy and hard samples, which are also diverse. To verify the efficacy of the proposed methods, extensive experiments are conducted using high and low-confidence samples separately. We observe from our experiments that the proposed hybrid sampling method, Least Confident and Diverse (LCD), consistently performs better compared to state-of-the-art methods. It is observed that selecting uncertain and diverse instances helps the model learn more distinct features. The codes related to this study will be available at https://github.com/XXX/LCD.

Multimodal Large Language Models (MLLMs) face significant computational overhead when processing long videos due to the massive number of visual tokens required. To improve efficiency, existing methods primarily reduce redundancy by pruning or merging tokens based on importance or similarity. However, these approaches largely overlook a critical dimension of video content, i.e., changes and turning points, and they lack a collaborative model for spatio-temporal relationships. To address this, we propose a new perspective: similarity is for identifying redundancy, while difference is for capturing key events. Based on this, we designed a training-free framework named ST-SimDiff. We first construct a spatio-temporal graph from the visual tokens to uniformly model their complex associations. Subsequently, we employ a parallel dual-selection strategy: 1) similarity-based selection uses community detection to retain representative tokens, compressing static information; 2) temporal difference-based selection precisely locates content-changing points to preserve tokens that capture key dynamic shifts. This allows it to preserve both static and dynamic content with a minimal number of tokens. Extensive experiments show our method significantly outperforms state-of-the-art approaches while substantially reducing computational costs. Our code is available in https://github.com/bingjunluo/ST-SimDiff.

High-resolution remote sensing images (RSIs) are crucial for Earth observation applications, yet acquiring them is often limited by sensor constraints and costs. In recent years, generative super-resolution (SR) methods, particularly diffusion models, have made significant progress. However, they typically require slow iterative inference with 40--1000 steps and exhibit limited flexibility in continuous-scale SR settings. To address these issues, we propose FlowGS, a generative reconstruction framework for arbitrary-scale SR of RSIs. FlowGS models the high-frequency detail representations between high- and low-resolution images and learns a continuous probability flow from noise to detail priors via flow matching (FM) constrained by shortcut consistency, thereby reducing generative complexity and improving inference efficiency. Additionally, we employ 2D Gaussian splatting to construct a continuous feature field, thereby enabling flexible reconstruction at arbitrary query locations. Experimental results show that FlowGS delivers competitive perceptual quality compared with existing methods in both continuous-scale and fixed-scale SR settings, with substantially improved inference efficiency.

Existing approaches for digital short-drama production typically rely on one-shot LLM generated scripts and loosely coupled pipelines, which fail to satisfy three key requirements of short-drama generation: (1) narrative pacing, resulting in weak hooks, insufficient escalation, and unattractive endings; (2) spatial consistency, leading to drifting scene layouts and inconsistent character positions across clips; and (3) production-level quality control, requiring extensive manual review and correction across script and visual stages. We present One Sentence, One Drama, a hierarchical multi-agent framework that transforms a user's single-sentence idea into a fully produced short drama through structured intermediate modules and iterative refinement. Our approach is built upon three key components: (1) a multi-agent debate-based story generation module that enforces short-drama pacing and narrative coherence; (2) a 3D-grounded first-frame generation mechanism that establishes a shared spatial reference for consistent character positioning and scene layout across clips; and (3) multi-stage reviewer loops that perform comprehensive error detection and targeted revision across script, visual, and video generation stages. We also introduce scene-level BGM matching and scene transition planning to improve the audience's immersive experience. To systematically evaluate this task, we introduce Short-Drama-Bench, a benchmark that extends standard video quality metrics with short-drama-specific criteria. Experimental results demonstrate that our method significantly outperforms existing pipelines in narrative quality, cross-clip consistency, and overall viewing experience.

Gait recognition enables non-intrusive, privacy-preserving identification but suffers in uncontrolled environments due to illumination and motion sensitivity of conventional cameras. In this work, we explore gait recognition using event cameras, which offer microsecond temporal resolution and high dynamic range, naturally capturing robust dynamic cues and suppressing static noise. Existing event-based approaches typically aggregate event streams into event images over long time windows, thereby discarding fine-grained motion dynamics critical for gait recognition. Therefore, we propose \textbf{EventGait}, an end-to-end dual-stream framework that separately models motion and shape while preserving the advantages of events. Our dynamic stream leverages a Mixture of Spiking Experts (MoSE) with diverse neuron constants for robust dynamic perception across complex motion and illumination scenes, while the static stream learns dense shape representations via Cross-modal Structure Alignment (CroSA) with large vision foundation models. To address the absence of large-scale event-based gait datasets, we introduce a synthesis pipeline and release two new benchmarks: SUSTech1K-E and CCGR-Mini-E. Extensive experiments have shown that event-based gait recognition not only achieves results comparable to camera-based gait recognition under normal conditions but also significantly outperforms it in low-light scenarios. Our approach sets a new state of the art on both synthesized and real-world event-based gait benchmarks, highlighting the robustness and potential of event-driven gait analysis. The code and datasets are released at https://github.com/QUEAHREN/EventGait.

Pretrained vision foundation models deliver strong performance across tasks with limited fine-tuning. However, their Vision Transformer (ViT) backbones impose high inference costs, limiting deployment on resource-constrained devices. In this work, we accelerate large-scale pretrained ViTs while preserving their feature extraction capabilities by exploiting the intrinsic convolution-like behavior of some attention heads. Specifically, we introduce an efficient depthwise convolution-based layer that serves as a drop-in replacement for these heads. Additionally, we propose simple strategies to identify which heads can be replaced and introduce a fine-tuning procedure that recovers downstream task performance. Across both image classification and segmentation tasks, our method achieves 17-20\% percent inference speedup with minimal performance degradation. We validate the approach through detailed derivations, extensive experiments, and efficiency benchmarks. The reference implementation is publicly available.

In everyday photography, aesthetically appealing moments are often captured with structural flaws (e.g., composition, camera viewpoint, or pose) that existing retouching and portrait enhancement methods cannot fix. We formulate Aesthetic Photo Reconstruction (APR) as improving a photo's aesthetic quality via structural reconstruction while preserving subject identity and scene semantics. Although recent advances in image editing models make APR feasible, they often lack aesthetic understanding, yielding edits that are semantically plausible yet aesthetically weak. To address this, we propose AesFormer, a two-stage framework that decouples aesthetic planning from image editing. In Stage 1, an aesthetic action model (AesThinker) analyzes the input along seven progressive photographic dimensions and outputs executable editing actions; we further apply GRPO-A to encourage broad exploration over diverse action plans beyond SFT. In Stage 2, an action-conditioned editor (AesEditor) performs structural edits guided by these actions. To support APR, we build a video-based corpus-mining pipeline (VCMP) and construct AesRecon, a benchmark of 9,071 strictly aligned (poor, good) image pairs. Experiments show that AesFormer substantially improves APR performance and is competitive with Nano Banana Pro. Code is available at https://github.com/PKU-ICST-MIPL/AesFormer_ICML2026.

Magnetic resonance imaging (MRI) is highly susceptible to patient motion due to its relatively long acquisition times and the fact that data are acquired sequentially in k-space. Even small patient movements introduce phase inconsistencies across measurements, leading to severe artifacts such as blurring, ghosting, and geometric distortions that can compromise diagnostic quality. Retrospective motion compensation remains challenging, particularly in accelerated acquisitions, due to the ill-posed nature of the joint reconstruction and motion estimation problem. In this work, we propose a unified Bayesian framework for motion-compensated 3D MRI that jointly estimates the anatomical image, rigid-body motion parameters, and coil sensitivity maps directly from motion-corrupted k-space data. Our approach integrates pretrained 3D complex-valued score-based diffusion models as expressive anatomical image priors within a physics-based forward model. Inference is performed by alternating diffusion posterior image updates with efficient proximal optimization steps for motion and coil sensitivity estimation, enabling fully unsupervised reconstruction without the need for paired motion-free training data. Experiments on simulated and real-motion brain MRI datasets demonstrate that the proposed method achieves improved image quality and motion robustness compared to state-of-the-art classical and learning-based motion correction techniques, particularly in the presence of severe motion and high acceleration.

Multimodal Large Language Models (MLLMs) are increasingly deployed in human-facing roles where personality perception is critical, yet existing benchmarks evaluate this capability solely on numerical Big Five score prediction, leaving open whether models truly perceive personality through behavioral understanding or merely prejudge through superficial pattern matching. We address this gap with three contributions. (i) A new task: we formalize Grounded Personality Reasoning (GPR), which requires MLLMs to anchor each Big Five rating in observable evidence through a chain of rating, reasoning, and grounding. (ii) A new dataset: we release MM-OCEAN (1,104 videos, 5,320 MCQs), produced by a multi-agent pipeline with human verification, with timestamped behavioral observations, evidence-grounded trait analyses, and seven categories of cue-grounding MCQs. (iii) Benchmark and analysis: we design a three-tier evaluation (rating, reasoning, grounding) plus four sample-level failure-mode metrics: Prejudice Rate (PR), Confabulation Rate (CR), Integration-failure Rate (IR), and Holistic-grounding Rate (HR), and benchmark 27 MLLMs (13 closed, 14 open). The analysis uncovers a striking Prejudice Gap: across the field, 51% of correct ratings are not grounded in retrieved cues, and the Holistic-Grounding Rate spans only 0-33.5%. These findings expose a disconnect between getting the right score and reasoning for the right reason, charting a roadmap for grounded social cognition in MLLMs.

Real-world image restoration is challenging due to complex and interacting mixed degradations. Recent agent-based approaches address this problem by composing multiple task-specific restoration tools. However, empirical analysis reveals that their performance is fundamentally limited by implicitly constrained planning spaces and the lack of coordination among independently pretrained tools. To address these issues, we propose OPERA (Optimized Planning-Execution Restoration Agent), a framework that jointly optimizes restoration planning and tool execution in an end-to-end manner. On the planning side, OPERA uses reinforcement learning to directly optimize tool composition over a combinatorial plan space, with the final restoration quality as the reward. On the execution side, OPERA introduces agent-guided co-training of restoration tools, enabling them to learn cooperative behaviors under sequential composition. Extensive experiments on multi-degradation benchmarks and real-world datasets demonstrate that OPERA consistently outperforms both all-in-one restoration models and existing agent-based methods across diverse and complex degradation scenarios.

The platonic representation hypothesis suggests that sufficiently large models converge to a shared representation geometry, even across modalities. Motivated by this, we ask: Can the semantic knowledge of a language model efficiently improve a vision model? As an answer, we introduce TextTeacher, a simple auxiliary objective that injects text embeddings as additional information into image classification training. TextTeacher uses readily available image captions, a pre-trained and frozen text encoder, and a lightweight projection to produce semantic anchors that efficiently guide representations during training while leaving the inference-time model unchanged. On ImageNet with standard ViT backbones, TextTeacher improves accuracy by up to +2.7 percentage points (p.p.) and yields consistent transfer gains (on average +1.0 p.p.) under the same recipe and compute. It outperforms vision knowledge distillation, yielding more accuracy at a constant compute budget or similar accuracy, but 33% faster. Our analysis indicates that TextTeacher acts as a feature-space preconditioner, shaping deeper layers in the first stages of training, and aiding generalization by supplying complementary semantic cues. TextTeacher adds negligible overhead, requires no costly multimodal training of the target model and preserves the simplicity and latency of pure vision models. Project page with code and captions: https://nauen-it.de/publications/text-teacher

Capsule endoscopy event detection is challenging because clinically relevant findings are sparse, visually heterogeneous, and evaluated at the event level rather than by frame accuracy. We propose VISTA, a metric-aligned multi-backbone framework for the RAREVISION task. VISTA combines EndoFM-LV for temporal context and DINOv3 ViTL/16 for frame-level visual semantics, followed by a Diverse Head Ensemble (DHE), Validation-Guided Weighted Fusion (VGWF), and Anatomy-Aware Temporal Event Decoding (ATED). The original official submission achieved hidden-test temporal mAP@0.5 of 0.3530 and mAP@0.95 of 0.3235. After the competition, extending local threshold refinement with a global coarse search improved performance to 0.3726 mAP@0.5 and 0.3431 mAP@0.95, ranking Team ACVLab second in the post-competition evaluation.

Vision-Language-Action (VLA) models have emerged as a promising framework for end-to-end autonomous driving. However, existing VLAs typically rely on sparse action supervision, which underutilizes their powerful scene understanding and reasoning capabilities. Recent attempts to incorporate dense visual supervision via world modeling often overemphasize pixel-level image reconstruction, neglecting semantically meaningful scene representation learning. In this work, we propose LVDrive, a Latent Visual representation enhanced VLA framework for autonomous driving. LVDrive introduces a future scene prediction task into the VLA paradigm, where future representations are learned entirely in a high-level latent space under auxiliary supervision from a pretrained vision backbone. Departing from inefficient autoregressive generation, we jointly model future scene and motion prediction within a unified embedding space, processed in a single forward pass to conduct the future-aware reasoning. We further design a two-stage trajectory decoding strategy that explicitly leverages the learned latent future representations to refine trajectory generation. Extensive experiments on the challenging Bench2Drive benchmark demonstrate that LVDrive achieves significant improvements in closed-loop driving performance, outperforming both action supervised methods and image-reconstruction-based world model approaches.

Driving Vision-Language-Action Models (Driving VLAs) commonly introduce natural-language reasoning as an intermediate interface for end-to-end planning, but reasoning-centric interfaces face three practical bottlenecks: obtaining high-quality reasoning annotations is difficult, generating and understanding long reasoning chains is challenging for compact models, and inference latency is substantially increased. In this paper, we rethink the design of language interfaces in Driving VLAs and show that concise one-step meta-actions are a simple yet effective alternative to verbose reasoning. Meta-actions provide semantic decision grounding while remaining low-entropy, and being automatically derivable from expert trajectories, enabling scalable supervision and reliable trajectory conditioning. Building on this interface, we propose DriveMA, which combines action-centric supervised training with a turn-level credit-assignment reinforcement learning framework that jointly optimizes meta-action correctness, trajectory quality, and trajectory--meta-action consistency. Experiments show that DriveMA already achieves a new state of the art on the Waymo End-to-End Driving Challenge with a 2B model, reaching a Rater Feedback Score (RFS) of 8.060, while its 4B version further improves the state of the art to 8.079; DriveMA also obtains competitive performance on NAVSIM. Ablations demonstrate that one-step meta-actions offer a better practical trade-off between expressiveness, predictability, and inference efficiency than natural-language reasoning or finer-grained action sequences. Code, data, and models will be released to facilitate future research.

Passive acoustic monitoring enables continuous, non-invasive biodiversity assessment across diverse ecosystems. The scale of these datasets has driven the adoption of machine learning, with supervised approaches showing strong performance. However, supervised methods require time-resolved annotated datasets, which remain scarce, especially in complex tropical soundscapes. We present PteroSet, a curated dataset of strongly annotated Neotropical bird vocalizations recorded in Puerto Asis (Putumayo) and Pivijay (Magdalena), Colombia, between 2023 and 2025. The dataset comprises 563 recordings (73.62 h) and 15,372 time-frequency annotations, including 6,702 events identified to the species level across 168 species. We release the annotations in a COCO-inspired JSON schema that unifies audio files, taxonomic categories, and labels for machine learning workflows. Beyond providing annotated data, PteroSet serves as a realistic benchmark that highlights key characteristics of tropical soundscapes, including acoustic co-occurrence and domain shift across recording sites. We provide a deep learning baseline for binary bird detection, demonstrating PteroSet's usability and the challenges it presents.

Training large multimodal models (LMMs) via reinforcement learning (RL) to natively invoke video-processing tools (e.g., cropping) has become a promising route to long-video understanding. However, existing native-RL methods dispatch tool calls sequentially (i.e., one per turn): a single wrong crop propagates errors without peer correction, multi-turn tool calls corrupt context, and inference cost scales linearly with the number of turns. We introduce ParaVT, the first multi-agent end-to-end RL-trained framework for Parallel Video Tool calling, dispatching multiple time-window crops in a single turn for cleaner context and better fault tolerance. Yet applying standard RL to ParaVT reveals an obstacle we term the Tool Prior Paradox: the pretrained tool priors that enable tool exploration also destabilize cold-started structural format and expose the skip-tool reward shortcut under temperature sampling. A cross-model contrast on a weaker-prior LMM supports this claim: format stays stable but RL elicits zero tool calls, indicating that prior strength is the shared driver of both format collapse and tool exploration. We propose PARA-GRPO (Parseability-Anchored and Ratio-gAted GRPO), which augments standard RL with two complementary mechanisms: (i) a targeted format reward applied only at the structural-token positions most prone to collapse, and (ii) a per-prompt frame-budget randomization that creates training prompts where calling the tool yields a measurable reward signal over skipping it. Across six long-video understanding benchmarks, ParaVT improves over the Qwen3-VL baseline by +7.9% on average, with PARA-GRPO lifting training-time format compliance from 0.13 to 0.64. As tool capabilities become increasingly internalized in modern LMMs, RL must cooperate with the resulting priors, and ParaVT offers a general recipe for agentic RL. Code, data, and model weights are publicly available.

RGB camera-based surveillance systems enable human action recognition for public safety and healthcare, yet raise serious privacy concerns. Existing methods rely on post-capture algorithms, which fail to protect privacy during data acquisition. We propose Lens Privacy Sealing (LPS), a simple hardware solution that physically obscures camera lenses with adjustable laminating film, providing pre-sensor privacy protection at minimal cost. Unlike software methods or expensive engineered optics, LPS achieves strong privacy through stochastic multi-layer scattering that is physically irreversible. We introduce the P$^3$AR dataset for privacy-preserving action recognition, featuring both large-scale replay-captured (P$^3$AR-NTU, 114K videos) and real-world collected (P$^3$AR-PKU) subsets with privacy attribute annotations. To handle video degradation from LPS, we propose MSPNet, a single-stage framework incorporating Inter-Frame Noise Suppressor (IFNS) and Cross-Frame Semantic Aggregator (CFSA), enhanced by contrastive language-image pre-training for robust semantic extraction. Extensive experiments demonstrate that MSPNet with IFNS and CFSA nearly doubles action recognition accuracy compared to baseline methods while suppressing identity recognition to low levels. Comprehensive validation shows LPS achieves a superior privacy-utility trade-off compared to state-of-the-art hardware methods, resists reconstruction attacks including PSF inversion and data-driven recovery, and generalizes robustly across optical configurations and challenging environments. Code is available at https://github.com/wangzy01/MSPNet.

MRI reconstruction is an inherently ill-posed inverse problem, since incomplete measurements admit many plausible solutions. This ambiguity becomes more severe under high acceleration, where pixel-domain continuous predictors tend to average over feasible reconstructions and suppress high-frequency anatomy. We address this limitation by moving reconstruction to discrete multi-scale latent space and posing it as autoregressive next-acceleration-scale prediction. Leveraging discrete priors proven effective in visual autoregressive modeling, our method restricts the solution to compact sequences of codebook tokens, enabling sharp reconstructions even from extremely sparse measurements. This discrete autoregressive formulation also aligns naturally with modern large language model post-training techniques. Building on this observation, we introduce on-policy privileged information distillation for visual autoregressive modeling, where a teacher is provided training only privileged context that is unavailable at inference, in our case fully sampled acquisitions, and supervises a student trained on its own rollouts, leading to consistent reconstruction gains. Through extensive experiments on the fastMRI benchmark, we show that our approach delivers improved reconstruction performance across diverse sampling patterns under extreme undersampling. Project website is \href{https://yilmazkorkmaz1.github.io/discrete-mri-reconstruction-opd/}{here}.

Conventional vision-language models (VLMs) struggle to interpret scenes captured under adverse conditions (e.g., low light, high dynamic range, or fast motion) because standard RGB images degrade in such environments. Event cameras provide a complementary modality: they asynchronously record per-pixel brightness changes with high temporal resolution and wide dynamic range, preserving motion cues where frames fail. We propose RE-VLM, the first dual-stream vision-language model that jointly leverages RGB images and event streams for robust scene understanding across both normal and challenging conditions. RE-VLM employs parallel RGB and event encoders together with a progressive training strategy that aligns heterogeneous visual features with language. To address the scarcity of RGB-Event-Text supervision, we further propose a graph-driven pipeline that converts synchronized RGB-Event streams into verifiable scene graphs, from which we synthesize captions and question-answer (QA) pairs. To develop and evaluate RE-VLM, we construct two datasets: PEOD-Chat, targeting illumination-challenged scenes, and RGBE-Chat, covering diverse scenarios. On captioning and VQA benchmarks, RE-VLM consistently outperforms state-of-the-art RGB-only and event-only models with comparable parameter counts, with particularly large gains under challenging conditions. These results demonstrate the effectiveness of event-augmented VLMs in achieving robust vision-language understanding across a wide range of real-world environments.

Due to the difficulty of obtaining ground-truth data for 4D radar scene flow estimation, previous methods typically rely on either self-supervised losses or cross-modal supervision using 3D LiDAR data, 2D images, and odometry. However, self-supervised approaches often yield suboptimal results due to radar's inherently low-fidelity measurements, while existing cross-modal supervised methods introduce complex multi-task architecture and require costly LiDAR sensors to generate pseudo radar scene flow labels from pretrained 3D tracking models. To overcome these limitations, we propose a task-specific iterative framework for weakly supervised radar scene flow learning, using only images and odometry for auxiliary supervision during training. Specially, we establish two novel instance-aware self-supervised losses by exploiting off-the-shelf 2D tracking and segmentation algorithms to obtain tracked instance masks, which are back-projected into 3D space to provide instance-level semantic guidance; for static regions, we integrate vehicle odometry with radar's intrinsic motion cues to construct a rigid static loss. Extensive experiments on the real-world View-of-Delft (VoD) dataset demonstrate that our method not only surpasses state-of-the-art cross-modal supervised approaches that rely on 3D multi-object tracking on dense LiDAR point clouds but also outperforms existing fully supervised scene flow estimation methods. The code is open-sourced at \href{https://github.com/FuJingyun/IterFlow}{https://github.com/FuJingyun/IterFlow}.

Achieving pixel-level accurate segmentation of structural cracks across diverse scenarios remains a formidable challenge. Existing methods face significant bottlenecks in balancing crack topology modeling with computational efficiency, often failing to reconcile high segmentation quality with low resource demands. To address these limitations, we propose the Ultra-Compact Structure-Calibrated Vision RWKV (SCRWKV), a network that achieves high-precision modeling via a novel Structure-Field Encoder (SFE) backbone while maintaining linear complexity. The SFE integrates the Adaptive Multi-scale Cascaded Modulator (AMCM) to enhance texture representation and utilizes the Structure-Calibrated Insight Unit (SCIU) as its core engine. Specifically, the SCIU employs the Geometry-guided Bidirectional Structure Transformation (GBST) to capture topological correlations and integrates the Dynamic Self-Calibrating Decay (DSCD) into Dy-WKV to suppress noise propagation. Furthermore, we introduce a lightweight Cross-Scale Harmonic Fusion (CSHF) decoder to achieve precise feature aggregation. Systematic evaluations on multiple benchmarks characterized by complex textures and severe interference demonstrate that SCRWKV, with only 1.22M parameters, significantly outperforms SOTA methods. Achieving an F1 score of 0.8428 and mIoU of 0.8512 on the TUT dataset, the model confirms its robust potential for efficient real-world deployment. The code is available at https://github.com/zhxhzy/SCRWKV.

Zero-shot composed image retrieval (ZS-CIR) retrieves a target image from a reference image and a text modification without human-annotated CIR triplets. Projection-based ZS-CIR methods are attractive because they do not rely on LLMs at inference and remain lightweight, but they often underperform LLM-based approaches on complex semantic modifications. This gap reflects a semantic transition bottleneck in projection-based ZS-CIR: endpoint-level matching can let the edit text act as a target-side attribute cue rather than grounding it as a source-conditioned semantic transition. We further show that adding semantic transition supervision to the same text adapter creates an endpoint--transition conflict between endpoint alignment and semantic transition alignment. To address this conflict, DeCIR decouples endpoint and transition learning. It constructs paired forward/reverse edit tuples from image-caption pairs, trains separate low-rank text adapter branches for endpoint alignment and semantic transition alignment, and merges them with Low-Rank Directional Merge (LRDM) into one deployable adapter. Extensive experiments on CIRR, CIRCO, FashionIQ, and GeneCIS demonstrate that DeCIR consistently improves projection-based ZS-CIR without increasing inference complexity.

Dense 3D reconstruction from continuous image streams requires both accurate geometric aggregation and stable long-term memory management. Recent feed-forward reconstruction frameworks integrate observations through persistent memory representations, yet most rely primarily on appearance-based similarity when updating memory. Such appearance-driven integration often leads to redundant accumulation of observations and unstable geometry when viewpoint changes occur. In this work, we propose a ray-aware pointer memory for streaming 3D reconstruction that explicitly models both spatial location and viewing direction within a unified memory representation. Each memory pointer stores its 3D position, associated ray direction, and feature embedding, allowing the system to reason jointly about geometric proximity and viewpoint consistency. Based on this representation, we introduce an adaptive pointer update strategy that replaces traditional fusion-based memory compression with a retain-or-replace mechanism. Instead of averaging nearby observations, the system selectively retains informative pointers while discarding redundant ones, preserving distinctive geometric structures while maintaining bounded memory growth. Furthermore, the joint reasoning over spatial distance and ray-direction discrepancy enables the system to distinguish between local redundancy, novel observations, and potential loop revisits in a unified manner. When loop candidates are detected, pose refinement is triggered to enforce global geometric consistency across the reconstruction. Extensive experiments demonstrate that the proposed ray-aware memory design significantly improves long-term reconstruction stability and camera pose accuracy while maintaining efficient streaming inference. Our approach provides a principled framework for scalable and drift-resistant online 3D reconstruction from image streams.

Accurately recovering human pose and appearance from video is an essential component of scene reconstruction, with applications to motion capture, motion prediction, virtual reality, and digital twinning. Despite significant interest in building realistic human avatars from video, this paper demonstrates that existing methods do not accurately recover the 3D geometry of humans. ViT-based approaches are not consistently reliable and can overfit to 2D views, while NeRF- and Gaussian Splatting-based avatars treat pose and appearance separately, limiting rendering generalization to new poses. To resolve these shortcomings, this paper proposes HumanSplatHMR, a joint optimization framework that refines 3D human poses while simultaneously learning a high-fidelity avatar for novel-view and novel-pose synthesis. Our key insight is to close the loop between geometric pose estimation and differentiable rendering. Unlike prior human avatar methods that rely on accurate human pose obtained through motion capture systems or offline refinement, which are impractical in in-the-wild scenarios, our approach uses only human mesh estimates from a state-of-the-art human pose estimator to better reflect real-world conditions. Therefore, instead of using the human pose only as a deformation prior, HumanSplatHMR backpropagates photometric, segmentation, and depth losses through a differentiable renderer to the pose parameters and global position. This coupling refines the global 3D pose over time, improving accuracy and alignment while producing better renderings from novel views. Experiments show consistent improvements over pose recovery baselines that omit image-level refinement and avatar baselines that decouple pose estimation from avatar reconstruction.

Large-scale 3D point clouds can consist of hundreds of millions of points. Even after downsampling, these point clouds are too large for modern 3D neural networks. In order to develop a semantic understanding of the scene, the point clouds are divided into smaller subclouds that can be processed. Typically, this division is done using spherical crops, resulting in a loss of surrounding geometric context. To address this issue, we propose alternative methods that produce subclouds with larger crop sizes while maintaining a similar number of points. Specifically, we compare exponential, Gaussian, and linear cropping methods with the spherical method. We evaluated three 3D deep learning model architectures using multiple indoor and outdoor environment datasets. Our results demonstrate that altering the cropping strategy can enhance model performance, especially for large-scale outdoor scenes, yielding new state-of-the-art results. Code is available at https://github.com/mvg-inatech/point_cloud_cropping

DeepSeek-OCR leverages visual-text compression to reduce long-text processing costs and accelerate inference, yet visual tokens remain prone to redundant textual and structural information. Moreover, current token pruning methods for conventional vision-language models (VLMs) fail to preserve textual fidelity due to improper compression mechanisms. By analyzing the decoding process of DeepSeek-OCR, we find that a distinct two-stage reading trajectory: the model initially prioritizes the majority of high-norm tokens, then subsequently redistributes its attention to the remaining ones. Motivated by this insight, we propose RTPrune, a two-stage token pruning method tailored for DeepSeek-OCR. In the first stage, we prioritize high-norm visual tokens that capture salient textual and structural information. In the second stage, the remaining tokens are paired and merged based on optimal transport theory to achieve efficient feature aggregation. We further introduce a dynamic pruning ratio that adapts to token similarity and textual density for OCR tasks, enabling a better efficiency-accuracy trade-off. Extensive experiments demonstrate state-of-the-art performance, as evidenced by 99.47% accuracy and 1.23$\times$ faster prefill on OmniDocBench, achieved with 84.25% token retention when applied to DeepSeek-OCR-Large.

We introduce AEGIS, A holistic benchmark for Evaluating forensic analysis of AI-Generated academic ImageS. Compared to existing benchmarks, AEGIS features three key advances: (1) Domain-Specific Complexity: covering seven academic categories with 39 fine-grained subtypes, exposing intrinsic forensic difficulty, where even GPT-5.1 reaches 48.80% overall performance and expert models achieve only limited localization accuracy (IoU 30.09%); (2) Diverse Forgery Simulations: modeling four prevalent academic forgery strategies across 25 generative models, with 11 yielding average forensic accuracy below 50%, showing that forensics lag behind generative advances; and (3) Multi-Dimensional Forensic Evaluation: jointly assessing detection, reasoning, and localization, revealing complementary strengths between model families, with multimodal large language models (MLLMs) at 84.74% accuracy in textual artifact recognition and expert detectors peaking at 79.54% accuracy in binary authenticity detection. By evaluating 25 leading MLLMs, nine expert models, and one unified multimodal understanding and generation model, AEGIS serves as a diagnostic testbed exposing fundamental limitations in academic image forensics.

The rapid proliferation of Vision-Language Models (VLMs) is often framed as enabling unified multimodal knowledge discovery but rests on an under-examined assumption: that current VLMs faithfully synthesise multimodal data. We argue they often do not, and this gap reflects a trustworthiness problem in the dominant Vision Encoder-Projector-LLM paradigm. Rather than extracting grounded knowledge from visual inputs, state-of-the-art models frequently exhibit functional blindness, i.e., exploiting strong language priors to bypass severe visual representation bottlenecks. In this work, we challenge the conventional methodology of multimodal evaluation, which relies on data ablation or new dataset creation and therefore conflates dataset biases with architectural incapacity. We propose an information-theoretic departure: the Modality Translation Protocol, designed to quantify what we call the Expense of Seeing. By translating semantic payloads rather than ablating them, we formulate three novel metrics -- the Toll (ToS), Curse (CoS), and Fallacy (FoS) of Seeing -- culminating in the Semantic Sufficiency Criterion (SSC). Furthermore, we hypothesise a Divergence Law of Multimodal Scaling: as the underlying language engines scale to unprecedented reasoning capabilities, the penalty of the visual knowledge bottleneck may increase rather than diminish. We argue the community should move beyond "multimodal gain" as a primary evaluation target. By elevating the SSC from a passive diagnostic constraint to an active architectural blueprint, we provide a foundation for guiding the next generation of AI systems toward genuine multimodal reasoning.

As AI models grow more complex, explainability is essential for building trust, yet concept-based counterfactual methods still face a trade-off between expressivity and efficiency. Representing underlying concepts as atomic sets is fast but misses relational context, whereas full graph representations are more faithful but require solving the NP-hard Graph Edit Distance (GED) problem. We propose U-CECE, a unified, model-agnostic multi-resolution framework for conceptual counterfactual explanations that adapts to data regime and compute budget. U-CECE spans three levels of expressivity: atomic concepts for broad explanations, relational sets-of-sets for simple interactions, and structural graphs for full semantic structure. At the structural level, both a precision-oriented transductive mode based on supervised Graph Neural Networks (GNNs) and a scalable inductive mode based on unsupervised graph autoencoders (GAEs) are supported. Experiments on the structurally divergent CUB and Visual Genome datasets characterize the efficiency-expressivity trade-off across levels, while human surveys and LVLM-based evaluation show that the retrieved structural counterfactuals are semantically equivalent to, and often preferred over, exact GED-based ground-truth explanations.

We propose a geometric framework for longitudinal multi-parametric MRI analysis based on patient-specific energy modelling in sequence space. Rather than operating on images with spatial networks, each voxel is represented by its multi-sequence intensity vector ($T1$, $T1c$, $T2$, FLAIR, ADC), and a compact implicit neural representation is trained via denoising score matching to learn an energy function $E_θ(\mathbf{u})$ over $\mathbb{R}^d$ from a single baseline scan. The learned energy landscape provides a differential-geometric description of tissue regimes without segmentation labels. Local minima define tissue basins, gradient magnitude reflects proximity to regime boundaries, and Laplacian curvature characterises local constraint structure. Importantly, this baseline energy manifold is treated as a fixed geometric reference: it encodes the set of contrast combinations observed at diagnosis and is not retrained at follow-up. Longitudinal assessment is therefore formulated as evaluation of subsequent scans relative to this baseline geometry. Rather than comparing anatomical segmentations, we analyse how the distribution of MRI sequence vectors evolves under the baseline energy function. In a paediatric case with later recurrence, follow-up scans show progressive deviation in energy and directional displacement in sequence space toward the baseline tumour-associated regime before clear radiological reappearance. In a case with stable disease, voxel distributions remain confined to established low-energy basins without systematic drift. The presented cases serve as proof-of-concept that patient-specific energy manifolds can function as geometric reference systems for longitudinal mpMRI analysis without explicit segmentation or supervised classification, providing a foundation for further investigation of manifold-based tissue-at-risk tracking in neuro-oncology.

Multimodal large language models (MLLMs) exhibit strong visual-language reasoning, yet cannot process structured, non-visual data such as human skeletons. Existing methods either compress skeleton dynamics into lossy feature vectors for text alignment, or quantize motion into discrete tokens that generalize poorly across heterogeneous skeleton formats. We present SkeletonLLM, which achieves universal skeleton understanding by translating arbitrary skeleton sequences into the MLLM's native visual modality. At its core is DrAction, a differentiable, format-agnostic renderer that converts skeletal kinematics into compact image sequences. Because the pipeline is end-to-end differentiable, MLLM gradients can directly guide the rendering to produce task-informative visual tokens. To further enhance reasoning capabilities, we introduce a cooperative training strategy: Causal Reasoning Distillation transfers structured, step-by-step reasoning from a teacher model, while Discriminative Finetuning sharpens decision boundaries between confusable actions. SkeletonLLM demonstrates strong generalization \revise{in open-vocabulary action recognition, while its learned reasoning capabilities naturally extend to motion captioning and question answering across heterogeneous skeleton formats} -- suggesting a viable path for applying MLLMs to non-native modalities. Code: https://github.com/wangzy01/SkeletonLLM.

Long-horizon action-conditioned video generation aims to synthesize temporally coherent videos that follow complex action instructions over extended horizons, requiring procedural ordering, persistent action execution, and scene consistency beyond conventional TI2V's short-term fidelity. Existing single-shot video generation models typically operate in an open-loop manner, leading to incomplete action execution, hallucinated motions, and temporal drift. To address this, we propose SPIRAL, a closed-loop framework that performs sequential planning and iterative reflection for action-conditioned long-horizon video generation. Specifically, SPIRAL instantiates a think-act-reflect process: a PlanAgent decomposes high-level goals into sub-actions, which condition a VideoGenerator to synthesize each segment alongside a memory context, while a CriticAgent evaluates intermediate video segments to provide corrective feedback for iterative refinement. This closed-loop design further supports self-evolution by utilizing PlanAgent-proposed actions and CriticAgent-derived rewards for GRPO-based post-training to enhance the video generator's long-horizon consistency. Moreover, we introduce ActVideoGen-Dataset for task-specific training, and establish ActVideoGen-Bench as a dedicated evaluation suite for measuring action quality and temporal coherence. Experiments across multiple TI2V backbones alongside the self-evolving strategy show consistent gains on ActVideoGen-Bench and VBench, demonstrating the effectiveness of SPIRAL.

Automated identification of DICOM image series is essential for large-scale medical image analysis, quality control, protocol harmonization, and reliable downstream processing. However, DICOM series classification remains challenging due to heterogeneous slice content, variable series length, and entirely missing, incomplete or inconsistent DICOM metadata. We propose an end-to-end multimodal framework for DICOM series classification that jointly models image content and acquisition metadata while explicitly accounting for all these challenges. (i) Images and metadata are encoded with modality-aware modules and fused using a bi-directional cross-modal attention mechanism. (ii) Metadata is processed by a sparse, missingness-aware encoder based on learnable feature dictionaries and value-conditioned modulation. By design, the approach does not require any form of imputation. (iii) Variability in series length and image data dimensions is handled via a 2.5D visual encoder and attention operating on equidistantly sampled slices. We evaluate the proposed approach on the publicly available Duke Liver MRI dataset and a large multi-institutional in-house cohort, assessing both in-domain performance and out-of-domain generalization. Across all evaluation settings, the proposed method consistently outperforms relevant image only, metadata-only and multimodal 2D/3D baselines. The results demonstrate that explicitly modeling metadata sparsity and cross-modal interactions improves robustness for DICOM series classification.

Human action recognition plays an important role when developing intelligent interactions between humans and machines. While there is a lot of active research on improving the machine learning algorithms for skeleton-based action recognition, not much attention has been given to the quality of the input skeleton data itself. This work demonstrates that by making use of multiple camera views to triangulate more accurate 3D~skeletons, the performance of state-of-the-art action recognition models can be improved significantly. This suggests that the quality of the input data is currently a limiting factor for the performance of these models. Based on these results, it is argued that the cost-benefit ratio of using multiple cameras is very favorable in most practical use-cases, therefore future research in skeleton-based action recognition should consider multi-view applications as the standard setup.

Convolutional networks require extensive image annotation, which can be costly and time-consuming. Feature Learning from Image Markers (FLIM) tackles this challenge by estimating encoder filters (i.e., kernel weights) from user-drawn markers on discriminative regions of a few representative images without traditional optimization. Such an encoder combined with an adaptive decoder comprises a FLIM network fully trained without backpropagation. Prior research has demonstrated their effectiveness in Salient Object Detection (SOD), being significantly lighter than existing lightweight models. This study revisits FLIM SOD and introduces FLIM-Bag of Feature Points (FLIM-BoFP), a considerably faster filter estimation method. The previous approach, FLIM-Cluster, derives filters through patch clustering at each encoder's block, leading to computational overhead and reduced control over filter locations. FLIM-BoFP streamlines this process by performing a single clustering at the input block, creating a bag of feature points, and defining filters directly from mapped feature points across all blocks. The paper evaluates the benefits in efficiency, effectiveness, and generalization of FLIM-BoFP compared to FLIM-Cluster and other state-of-the-art baselines for parasite detection in optical microscopy images.

Augmented reality (AR) guidance in laparoscopic liver surgery requires accurate registration of preoperative 3D models to intraoperative 2D video, but remains challenging due to partial visibility, specularities, and tissue deformation. Existing methods often rely on contour-based rigid initialization and finite-element (FE) models for deformable registration, increasing modeling and engineering complexity. We present a depth-augmented, FE-free 3D--2D registration pipeline that combines robust rigid initialization with patient-specific non-rigid refinement. For rigid alignment, we adapt the RefineNet module of FoundationPose to laparoscopic liver scenes by using multi-class contour maps and monocular depth for relative pose refinement. For deformable alignment, we construct a patient-specific statistical deformation model from non-rigid ICP (NICP) correspondences and optimize pose and shape parameters using a coarse-to-fine L-BFGS-B strategy. On a public clinical laparoscopic liver dataset, the proposed method achieves a mean target registration error (TRE) of 14.73\,mm under a controlled manual-contour setting designed to isolate registration performance. Ablation studies show that monocular depth improves rigid initialization over contour-only inputs, while tumor-mapping analysis indicates that good surface alignment does not necessarily translate into lower target localization error. On an external dataset without ground truth, the method produces visually plausible overlays for qualitative assessment. These results suggest that depth-augmented pose refinement and FE-free statistical deformation modeling provide a promising alternative to FE-based pipelines for controlled 3D--2D liver registration in surgical AR.

Adapting large pre-trained models to downstream tasks often produces task-specific parameter updates that are expensive to relearn for every model variant. While recent work has shown that such updates can be transferred between models with identical architectures, transferring them across models of different widths remains unexplored. In this work, we introduce Theseus, a training-free method for transporting task updates across heterogeneous-width models. Rather than matching parameters, we characterize a task update by the functional effect it induces on intermediate representations. We formalize task-vector transport as a functional matching problem on observed activations and show that, after aligning representation spaces via orthogonal Procrustes analysis, it admits a stable closed-form solution that preserves the geometry of the update. We evaluate Theseus on vision and language models across different widths, showing consistent improvements over baselines without additional training or backpropagation. Our results show that task updates can be meaningfully transferred across architectures when task identity is defined functionally rather than parametrically. Code is available at https://github.com/apanariello4/merge-and-rebase.

This paper surveys the state-of-the-art in the nexus of SLAM and Wireless Communications, attributing the bidirectional impact of each with a focus on visual SLAM (V-SLAM) integration. We provide an overview of key concepts related to wireless signal propagation, geometric channel modeling, and radio frequency (RF)-based localization and sensing. In addition to this, we show image processing techniques that can detect landmarks, proactively predicting optimal paths for wireless channels. Several dimensions are considered, including the prerequisites, techniques, background, and future directions and challenges of the intersection between SLAM and wireless communications. We analyze estimation and control approaches such as Bayesian filters, feature-based pose estimation, perception-aware motion control, spatial methods for signal processing such as vector fields, and key technological aspects. We expose techniques and items towards enabling a highly effective retrieval of the autonomous robot state. Among other interesting findings, we observe that monocular V-SLAM would benefit from RF relevant information, as the latter can serve as a proxy for the scale ambiguity resolution. Conversely, we find that wireless communications in the context of 5G and beyond can potentially benefit from visual odometry that is central in SLAM. Moreover, we examine other sources besides the camera for SLAM and describe the twofold relation with wireless communications. Finally, integrated solutions performing joint communications and SLAM appear to be in their infancy: theoretical and practical advancements are required to add higher-level localization and semantic perception capabilities to RF and multi-antenna technologies.

Fake Image Detection (FID), aiming at unified detection across four image forensic subdomains, is critical in real-world forensic scenarios. Compared with ensemble approaches, monolithic FID models are theoretically more promising, but to date, consistently yield inferior performance in practice. In this work, we identify the intrinsic distinctness of artifacts across subdomains, a critical barrier we term the ``Ji-Zhe phenomenon". Driven by this phenomenon, we diagnose the cause of this underperformance for the first time: the collapse of the artifact feature space. The core challenge for developing a practical monolithic FID model thus boils down to the ``unified-yet-discriminative" reconstruction of the artifact feature space. To address this paradoxical challenge, we hypothesize that high-level semantics can serve as a structural prior for the reconstruction, and further propose Semantic-Induced Constrained Adaptation (SICA), the first monolithic FID paradigm. Extensive experiments on our OpenMMSec dataset demonstrate that SICA outperforms 15 state-of-the-art methods and reconstructs the target unified-yet-discriminative artifact feature space in a near-orthogonal manner, thus firmly validating our hypothesis. The code and dataset are available at: https://github.com/venus-guangjian/SICA_OpenMMSec.

Model merging combines multiple fine-tuned models into a single model by adding their weight updates, providing a lightweight alternative to retraining. Existing methods primarily target resolving conflicts between task updates, leaving the failure mode of over-counting shared knowledge unaddressed. We show that when tasks share aligned spectral directions (i.e., overlapping singular vectors), a simple linear combination repeatedly accumulates these directions, inflating the singular values and biasing the merged model toward shared subspaces. To mitigate this issue, we propose Singular Value Calibration (SVC), a training-free and data-free post-processing method that quantifies subspace overlap and rescales inflated singular values to restore a balanced spectrum. Across vision and language benchmarks, SVC consistently improves strong merging baselines and achieves state-of-the-art performance. Furthermore, by modifying only the singular values, SVC improves the performance of Task Arithmetic by 13.0%. Code is available at https://github.com/lyymuwu/SVC.

Vision tool-use reinforcement learning (RL) can equip vision language models with visual operators such as crop-and-zoom and achieves strong performance gains, yet it remains unclear whether these gains are driven by improvements in tool use or evolving intrinsic capabilities. We introduce MED (Measure--Explain--Diagnose), a coarse-to-fine framework that disentangles intrinsic capability changes from tool-induced effects, decomposes the tool-induced performance difference into gain and harm terms, and probes the mechanisms driving their evolution. Across checkpoint-level analyses in the crop-and-zoom setting on two VLMs with different tool priors and six benchmarks, we find that improvements are dominated by intrinsic learning, while tool-use RL mainly reduces tool-induced harm (e.g., fewer call-induced errors and weaker tool schema interference) and yields limited progress in tool-based correction of intrinsic failures. Overall, in the crop-and-zoom setting studied here, current vision tool-use RL learns to coexist safely with tools rather than master them.

Existing multimodal large language models for long-video understanding predominantly rely on uniform sampling and single-turn inference, limiting their ability to identify sparse yet critical evidence amid extensive redundancy. We introduce Video-o3, a novel framework that supports iterative discovery of salient visual clues, fine-grained inspection of key segments, and adaptive termination once sufficient evidence is acquired. Technically, we address two core challenges in interleaved tool invocation. First, to mitigate attention dispersion induced by the heterogeneity of reasoning and tool-calling, we propose Task-Decoupled Attention Masking, which isolates per-step concentration while preserving shared global context. Second, to control context length growth in multi-turn interactions, we introduce a Verifiable Trajectory-Guided Reward that balances exploration coverage with reasoning efficiency. To support training at scale, we further develop a data synthesis pipeline and construct Seeker-173K, comprising 173K high-quality tool-interaction trajectories for effective supervised and reinforcement learning. Extensive experiments show that Video-o3 substantially outperforms state-of-the-art methods, achieving 72.1% accuracy on MLVU and 46.5% on Video-Holmes. These results demonstrate Video-o3's strong multi-hop evidence-seeking and reasoning capabilities, and validate the effectiveness of native tool invocation in long-video scenarios.

Recent Vision-Language Models (e.g., ColPali) enable fine-grained Visual Document Retrieval (VDR) but incur prohibitive multi-vector index storage overhead. Existing training-free pruning methods either rely on heuristic layer choices or degrade sharply under aggressive compression, leading prior work to argue that effective high-compression pruning requires query-dependent training. We challenge this view with Structural Anchor Pruning (SAP), a self-calibrating, training-free, and query-agnostic index-time pruning framework with three components: (i) Score Retention (SR), a white-box per-layer compression diagnostic; (ii) SR-guided window selection, a procedure that automatically locates the structural pruning region for any backbone with no per-model hyperparameters; and (iii) a visual in-degree centrality scorer that identifies anchor patches within the selected window. On the ViDoRe v1/v2 benchmarks across three architectures spanning 18, 28, and 36 backbone layers, SAP retains over 90\% of NDCG@5 while pruning more than 90\% of visual tokens, without any per-model parameter tuning. Our layer-resolved SR analysis reveals an Alignment-Aggregation Divergence: the document's visual structure is preserved as a stable ``Structural Plateau'' within the backbone, but the final layers reshape this representation into a sparse, query-aligned form that is no longer suitable for pruning. This is the mechanistic reason SAP succeeds where final-layer methods fail.

We present UIKA, a feed-forward animatable Gaussian head model from an arbitrary number of pose-free inputs, including a single image, multi-view captures, and smartphone-captured videos. Unlike the traditional avatar method, which requires a studio-level multi-view capture system and reconstructs a human-specific model through a long-time optimization process, we rethink the task through the lenses of model representation, network design, and data preparation. First, we introduce a UV-guided avatar modeling strategy, in which each input image is associated with a pixel-wise facial correspondence estimation. Such correspondence estimation allows us to reproject each valid pixel color from screen space to UV space, which is independent of camera pose and character expression. Furthermore, we design learnable UV tokens on which the attention mechanism can be applied at both the screen and UV levels. The learned UV tokens can be decoded into canonical Gaussian attributes using aggregated UV information from all input views. To train our large avatar model, we additionally prepare a large-scale, identity-rich synthetic training dataset. Our method significantly outperforms existing approaches in both monocular and multi-view settings.

Single-view RGB model-based object pose estimation methods achieve strong generalization but are fundamentally limited by depth ambiguity, clutter, and occlusions. Multi-view pose estimation methods have the potential to solve these issues, but existing works rely on precise single-view pose estimates or lack generalization to unseen objects. We address these challenges via the following three contributions. First, we introduce AlignPose, a 6D object pose estimation method that aggregates information from multiple extrinsically calibrated RGB views and does not require any object-specific training or symmetry annotation. Second, the key component of this approach is a new multi-view feature-metric refinement specifically designed for object pose. It optimizes a single, consistent world-frame object pose by minimizing the feature discrepancy between on-the-fly rendered object features and observed image features across all views simultaneously. Third, we report extensive experiments on six datasets (YCB-V, T-LESS, HouseCat6D, ITODD-MV, IPD, XYZ-IBD) using the BOP benchmark evaluation and show that AlignPose outperforms other published methods, especially on challenging industrial datasets where multiple views are readily available in practice.

Large multimodal models (LMMs) have shown great potential for video reasoning with textual Chain-of-Thought. However, they remain vulnerable to hallucinations, especially when processing long-form videos where evidence is sparse and temporally dispersed. Inspired by how humans comprehend long videos - by first skimming globally and then examining relevant clips for details - we introduce LongVT, an end-to-end agentic framework that enables "Thinking with Long Videos" via interleaved Multimodal Chain-of-Tool-Thought. Specifically, we exploit LMMs' inherent temporal grounding ability as a native video cropping tool to zoom in on a specific video clip and resample finer-grained video frames. This global-to-local reasoning loop continues until answers are grounded in retrieved visual evidence. Given the scarcity of fine-grained question-answering (QA) data for the long video reasoning task, we curate and will release a data suite named VideoSIAH to facilitate both training and evaluation. Specifically, our training dataset consists of 247.9K samples for tool-integrated cold-start supervised fine-tuning, 1.6K samples for agentic reinforcement learning, and 15.4K samples for agentic reinforcement fine-tuning, respectively. Our evaluation benchmark consists of 1,280 QA pairs that are carefully curated through a semi-automatic data pipeline with human-in-the-loop validation. With a meticulously designed three-stage training strategy and extensive empirical validation, LongVT consistently outperforms existing strong baselines across four challenging long-video understanding and reasoning benchmarks. Our codes, data, and model checkpoints are publicly available at https://github.com/EvolvingLMMs-Lab/LongVT .

Despite the rise of billion-parameter foundation models trained across thousands of GPUs, similar scaling gains have not been shown for humanoid control. Current neural controllers for humanoids remain modest in size, target a limited set of behaviors, and are trained on a handful of GPUs. We show that scaling model capacity, data, and compute yields a generalist humanoid controller capable of natural, robust whole-body movements. We position motion tracking as a scalable task for humanoid control, leveraging dense supervision from diverse motion-capture data to acquire human motion priors without manual reward engineering. We build a foundation model for motion tracking by scaling along three axes: network size (1.2M to 42M parameters), dataset volume (100M+ frames from 700 hours of motion capture), and compute (21k GPU hours). Beyond demonstrating the benefits of scale, we further show downstream utility through: (1) a real-time kinematic planner bridging motion tracking to tasks such as navigation, enabling natural and interactive control, and (2) a unified token space supporting VR teleoperation and vision-language-action (VLA) models with a single policy. Through this interface, we demonstrate autonomous VLA-driven whole-body loco-manipulation requiring coordinated hand and foot placement. Scaling motion tracking exhibits favorable properties: performance improves steadily with compute and data diversity, and learned policies generalize to unseen motions, establishing motion tracking at scale as a practical foundation for humanoid control.

The detection of Protected Health Information (PHI) in medical imaging is critical for safeguarding patient privacy and ensuring compliance with regulatory frameworks. Traditional detection methodologies predominantly utilize Optical Character Recognition (OCR) models in conjunction with named entity recognition. However, recent advancements in Large Multimodal Model (LMM) present new opportunities for enhanced text extraction and semantic analysis. In this study, we systematically benchmark three prominent closed and open-sourced LMMs, namely GPT-4o, Gemini 2.5 Flash, and Qwen 2.5 7B, utilizing two distinct pipeline configurations: one dedicated to text analysis alone and another integrating both OCR and semantic analysis. Our results indicate that LMM exhibits superior OCR efficacy (WER: 0.03-0.05, CER: 0.02-0.03) compared to conventional models like EasyOCR. However, this improvement in OCR performance does not consistently correlate with enhanced overall PHI detection accuracy. The strongest performance gains are observed on test cases with complex imprint patterns. In scenarios where text regions are well readable with sufficient contrast, and strong LMMs are employed for text analysis after OCR, different pipeline configurations yield similar results. Furthermore, we provide empirically grounded recommendations for LMM selection tailored to specific operational constraints and propose a deployment strategy that leverages scalable and modular infrastructure.

Flow Matching (FM) underpins many state-of-the-art generative models, yet recent results indicate that Transition Matching (TM) can achieve higher quality with fewer sampling steps. This work answers the question of when and why TM outperforms FM. First, when the target is a unimodal Gaussian distribution, we prove that TM attains strictly lower KL divergence than FM for finite number of steps. The improvement arises from stochastic difference latent updates in TM, which preserve target covariance that deterministic FM underestimates. We then characterize convergence rates, showing that TM achieves faster convergence than FM under a fixed compute budget, establishing its advantage in the unimodal Gaussian setting. Second, we extend the analysis to Gaussian mixtures and identify local-unimodality regimes in which the sampling dynamics approximate the unimodal case, where TM can outperform FM. The approximation error decreases as the minimal distance between component means increases, highlighting that TM is favored when the modes are well separated. However, when the target variance approaches zero, each TM update converges to the FM update, and the performance advantage of TM diminishes. In summary, we show that TM outperforms FM when the target distribution has well-separated modes and non-negligible variances. We validate our theoretical results with controlled experiments on Gaussian distributions, and extend the comparison to real-world applications in image and video generation.

Recent video generation models can produce smooth and visually appealing clips, but they often struggle to synthesize complex dynamics with a coherent chain of consequences. Accurately modeling visual outcomes and state transitions over time remains a core challenge. In contrast, large language and multimodal models (e.g., GPT-4o) exhibit strong visual state reasoning and future prediction capabilities. To bridge these strengths, we introduce VChain, a novel inference-time chain-of-visual-thought framework that injects visual reasoning signals from multimodal models into video generation. Specifically, VChain contains a dedicated pipeline that leverages large multimodal models to generate a sparse set of critical keyframes as snapshots, which are then used to guide the sparse inference-time visual-state adaptation of a pre-trained video generator only at these key moments. Our approach is tuning-efficient, introduces minimal overhead and avoids dense supervision. Extensive experiments on complex, multi-step scenarios show that VChain significantly enhances the quality of generated videos.

Diffusion Transformers (DiTs) have recently emerged as a powerful backbone for image generation, demonstrating superior scalability and performance over U-Net architectures. However, their practical deployment is hindered by substantial computational and memory costs. While Quantization-Aware Training (QAT) has shown promise for U-Nets, its application to DiTs faces unique challenges, primarily due to the sensitivity and distributional complexity of activations. In this work, we identify activation quantization as the primary bottleneck for pushing DiTs to extremely low-bit settings. To address this, we propose a systematic QAT framework for DiTs, named RobuQ. We start by establishing a strong ternary weight (W1.58A4) DiT baseline. Building upon this, we propose RobustQuantizer to achieve robust activation quantization. Our theoretical analyses show that the Hadamard transform can convert unknown per-token distributions into per-token normal distributions, providing a strong foundation for this method. Furthermore, we propose AMPN, the first Activation-only Mixed-Precision Network pipeline for DiTs. This method applies ternary weights across the entire network while allocating different activation precisions to each layer to eliminate information bottlenecks. Through extensive experiments on unconditional and conditional image generation, our RobuQ framework achieves state-of-the-art performance for DiT quantization in sub-4-bit quantization configuration. To the best of our knowledge, RobuQ is the first achieving stable and competitive image generation on large datasets like ImageNet-1K with activations quantized to average 2 bits. The code and models will be available at https://github.com/racoonykc/RobuQ .

Generalized Category Discovery (GCD) aims to identify both known and novel categories within unlabeled data by leveraging a set of labeled examples from known categories. Existing GCD methods primarily depend on semantic labels and global image representations, often overlooking the detailed part-level cues that are crucial for distinguishing closely related categories. In this paper, we introduce PartCo, short for Part-Level Correspondence Prior, a novel framework that enhances category discovery by incorporating part-level visual feature correspondences. By leveraging part-level relationships, PartCo captures finer-grained semantic structures, enabling a more nuanced understanding of category relationships. Importantly, PartCo seamlessly integrates with existing GCD methods without requiring significant modifications. Our extensive experiments on multiple benchmark datasets demonstrate that PartCo significantly improves the performance of current GCD approaches, outperforming most existing methods by bridging the gap between semantic labels and part-level visual compositions, thereby setting new benchmarks for GCD.

Object detection in autonomous driving is frequently compromised by complex illumination. While event cameras offer a robust solution, they are susceptible to sudden contrast changes such as reflections which often trigger dense, misleading event signals. To overcome this, we leverage RGB-derived surface normal maps as explicit geometric constraints. Crucially, even when RGB degrades, they preserve low-frequency structural priors that effectively assist in event-based detection. Consequently, we present NRE-Net, a trimodal framework that integrates structural priors from surface Normal maps, appearance context from RGB images, and high-frequency dynamics from Events. The Adaptive Dual-stream Fusion Module (ADFM) first aligns geometric and appearance cues, followed by the Event-modality Aware Fusion Module (EAFM) which selectively integrates event dynamics. Extensive evaluations on DSEC-Det-sub and PKU-DAVIS-SOD demonstrate that incorporating geometric priors yields an additional 3.0% AP50 gain over dual-modal baselines, while our approach consistently outperforms fusion methods such as SFNet (+2.7%) and SODFormer (+7.1%).

Rotary Position Embedding (RoPE) is widely adopted in large language models, but when applied to vision-language models (VLMs) it couples text and image position indices and can introduce spurious cross-modal relative-position bias. We propose Per-Token Distance (PTD) to quantify cross-modal positional disentanglement, and prove that PTD = 0 is a sufficient condition to eliminate the geometric attention bias induced by RoPE. Guided by this criterion, we introduce Circle-RoPE, which remaps 2D image-token coordinates onto an annulus orthogonal to the text position axis, yielding a cone-like geometry where each text token is equidistant to all image tokens while preserving intra-image spatial structure. We further propose Alternating Geometry Encoding (AGE) to combine complementary geometric priors by alternating the decoupled geometry of Circle-RoPE and the grid-based prior of standard RoPE across layers. This design enables cross-modal positional disentanglement while preserving fine-grained intra-image spatial structure. Experiments on diverse VLM backbones and multimodal benchmarks show consistent gains in spatial grounding and visual reasoning. The code is available at https://github.com/lose4578/CircleRoPE.

The Leaf Area Index (LAI) is a critical parameter to understand ecosystem health and vegetation dynamics. In this paper, we propose a novel method for pixel-wise LAI prediction by leveraging the complementary information from Sentinel 1 radar data and Sentinel 2 multi-spectral data at multiple timestamps. Our approach uses a deep neural network based on multiple U-nets tailored specifically to this task. To handle the complexity of the different input modalities, it is comprised of several modules that are pre-trained separately to represent all input data in a common latent space. Then, we fine-tune them end-to-end with a common decoder that also takes into account seasonality, which we find to play an important role. Our method achieved 0.06 RMSE and 0.93 R2 score on publicly available data. We make our contributions available at https://github.com/valentingol/LeafNothingBehind for future works to further improve on our current progress.

Reliable people detection is crucial for the safe autonomy of mobile robots and heavy vehicles, both on roads and in industrial settings like mining and construction. However, common sensors like cameras or lidars are prone to failure in adverse conditions such as dust, fog, or smoke, which limits their use in real-world robotic systems. Radar, on the other hand, delivers robust measurements in a wide range of environmental conditions. In particular, modern high-resolution 4D imaging radars provide 4D point clouds across range, azimuth, and elevation, as well as per-point Doppler velocity data, well suited for robot perception. We propose TMVA4D, a family of artificial neural network architectures based on CNN and ConvLSTM encoders that leverage the 4D radar modality for semantic segmentation. The architectures are trained to distinguish between background and person classes using a series of 2D projections of the 4D radar data, encompassing elevation, azimuth, range, and Doppler velocity dimensions. Evaluated across several operational sites, our models achieve promising performance (Dice 75.9%, IoU 61.2% for class person) even in low-visibility conditions. The data and code will be made publicly available upon publication.

Spiking neural networks (SNNs) have attracted much attention for their high energy efficiency and recent advances in classification performance. However, unlike traditional deep learning approaches, the study of SNN robustness to adversarial examples remains relatively underdeveloped. In this work, we advance the adversarial attack side of SNNs through three contributions. First, we show that successful white-box adversarial attacks on SNNs are highly dependent on the underlying surrogate gradient estimator, even for adversarially trained SNNs. Second, using the best single surrogate gradient estimator, we analyze the transferability of adversarial attacks across SNNs, Vision Transformers (ViTs) and CNNs. Our analysis reveals two key gaps: no existing white-box attack exploits multiple surrogate gradient estimators for SNNs, and no single-model attack reliably generates adversarial examples that simultaneously fool both SNN and non-SNN models. For our third contribution, we develop the Mixed Dynamic Spiking Estimation (MDSE) attack to address these issues. MDSE uses a dynamic gradient estimation scheme to fully exploit multiple surrogate gradient estimator functions and generates adversarial examples capable of fooling SNN and non-SNN models simultaneously. MDSE is up to 91.4% more effective on SNN/ViT model ensembles and provides a 3x boost on adversarially trained SNN ensembles compared to conventional white-box attacks like Auto-PGD. Experiments cover three datasets (CIFAR-10, CIFAR-100, ImageNet) and nineteen classifier models (seven per CIFAR dataset, five for ImageNet). Our implementation of MDSE and the evaluated models is publicly available at https://github.com/nuoxuxxx/attacking-the-spike-mdse.

Modern vision models have achieved strong object-recognition performance, yet it remains unclear whether their representations encode object-level semantic relatedness, the meaningful connection between object concepts that supports human visual cognition. Existing benchmarks predominantly target category prediction or rely on image--text matching, leaving the visual representation itself underexamined. Drawing on cognitive psychology, we recast semantic relatedness as a triplet-ranking task and study two image-only test beds: POPORO, an existing 400-triplet psychological stimulus set repurposed for representation evaluation, and PoporoIN, a newly constructed and manually curated 1,000-triplet ImageNet-validation extension. Each triplet is annotated along two orthogonal axes: a related-target axis distinguishing Categorical Relatedness (CR, taxonomic) from conTextual Relatedness (TR, thematic), and a distractor axis distinguishing Color-matched Distractors (CD) from Shape-matched Distractors (SD). Twenty pretrained models spanning supervised, self-supervised, vision--language, and generative paradigms were evaluated by cosine similarity in an inference-only protocol. Transformer-based representations exceeded convolutional counterparts by up to 18.30 percentage points on PoporoIN at comparable ImageNet accuracy, and vision--language encoders exceeded vision-only counterparts by up to 22.50 percentage points under matched ImageNet accuracy on POPORO. Across paradigms, models recognized taxonomic targets more reliably than thematic ones and were more easily misled by shape-matched than by color-matched distractors. The benchmarks expose representational properties that classification accuracy alone does not fully predict, bridging cognitive psychology and visual representation evaluation.

Human Activity Recognition (HAR) has shown remarkable effectiveness in various applications, such as smart healthcare and intelligent manufacturing. However, a major challenge faced by HAR is the distribution shift across different sensor data domains, which often leads to decreased performance when deployed for real-world applications. To address this issue, this paper introduces GenHAR, a novel framework designed to mitigate the domain gap by learning domain-invariant sensor representations. GenHAR aims to enhance the generalization capabilities of HAR on target domains purely with data from the source domain. The key novelty of GenHAR lies in two aspects. Firstly, GenHAR tokenizes sensor data and learns correlations among frequency sensor channel dimensions to improve the robustness of HAR models. Secondly, GenHAR improves the efficiency via selective masking and an efficient attention mechanism. We conduct a systematic analysis of GenHAR by comparing it with state-of-the-art HAR methods on real-world human activity datasets. Results show that GenHAR outperforms state-of-the-art methods by 9.97% in accuracy, and reduces Floating Point Operations by 6.4 times. Moreover, we deploy GenHAR at a leading logistics company in 4 cities, and have detected 2.15 billion real-time activities. We release our code at: https://github.com/Sensor-FoundationModel/GenHAR.

Recent advances in Multimodal Large Language Models (MLLMs) have significantly advanced video understanding tasks, yet challenges remain in efficiently compressing visual tokens while preserving spatiotemporal interactions. Existing methods, such as LLaVA family, utilize simplistic pooling or interpolation techniques that overlook the intricate dynamics of visual tokens. To bridge this gap, we propose ST-GridPool, a novel training-free visual token enhancement method designed specifically for Video LLMs. Our approach integrates Pyramid Temporal Gridding (PTG), which captures multi-grained spatiotemporal interactions through hierarchical temporal gridding, and Norm-based Spatial Pooling (NSP), which preserves high-information visual regions by leveraging the correlation between token norms and semantic richness. Extensive experiments on various benchmarks demonstrate that ST-GridPool consistently enhances performance of Video LLMs without requiring costly retraining. Our method offers an efficient and plug-and-play solution for improving visual token representations. Our code is available in https://github.com/bingjunluo/ST-GridPool.

Reinforcement learning with verifiable rewards (RLVR) has emerged as a promising paradigm for advancing complex reasoning in large language models, and recent work extends RLVR to multimodal large language models (MLLMs). This transfer, however, surfaces a faithfulness challenge: faithful perception of task-relevant visual evidence and faithful use of that evidence during reasoning, leading to unsatisfactory gains on multimodal benchmarks. Specifically, existing perception supervision often operates on textual descriptions rather than natively on image regions, and faithful use is largely overlooked, exposing the perception-reasoning disconnect where correctly perceived evidence is dropped or contradicted during reasoning. To close these gaps, we propose Faithful-MR1, a training framework that anchors and reinforces visual attention to address both halves of faithful multimodal reasoning. The Anchoring stage turns perception into an explicit pre-reasoning subtask, supervising a dedicated <Focus> token's attention directly against image regions rather than through textual descriptions. The Reinforcing stage exposes faithful use through counterfactual image intervention, rewarding answer-correct trajectories that concentrate visual attention where vision causally matters. Extensive experiments demonstrate that Faithful-MR1 outperforms recent multimodal reasoning baselines on both Qwen2.5-VL-Instruct 3B and 7B backbones while using substantially less training data.

We formalize and enable the task of open tree decomposition, which segments an image into hierarchical trees of visual components with unconstrained granularity and flexibility. Specifically, we provide the foundation benchmark for this new paradigm with the following three key contributions. First, we overcome the prohibitively high cognitive and physical bottlenecks of manual annotation by developing a fully automated generation pipeline that synergizes the semantic reasoning of Large Vision-Language Models (LVLMs) with the precise geometric grounding of SAM 3. Second, leveraging this pipeline, we construct COCOTree, a massive-scale benchmark featuring over 21K images and 1.8M structural nodes. By embracing an open-vocabulary space of over 3.5K unique labels, it successfully captures the long-tail distribution of complex physical assemblies. Notably, rigorous human evaluation confirms our generated annotations demonstrate strong alignment with human structural judgment. Third, we establish a standardized evaluation protocol by proposing the Open Tree Quality (OTQ) metric, which jointly assesses mask precision, label accuracy, and structural consistency. We release our dataset and benchmark code at https://github.com/melonkick3090/COCOTree.

Reconstructing 4D (3D+t) cardiac geometry from sparse 2D echocardiography is highly desirable yet fundamentally challenged by geometric ambiguity and temporal discontinuity. To tackle these issues, we propose Echo4DIR, a novel test-time 4D implicit reconstruction framework. Specifically, we learn robust 3D shape priors from statistical shape models (SSMs) via a cardiac conditional SDF, constructing an Epipolar Mask Encoder module with epipolar cross attention to effectively fuse multi-view features. To bridge the synthetic-to-real domain gap, we introduce a self-supervised SDF-tailored differentiable rendering strategy for patient-specific 3D shape adaptation using uncalibrated clinical masks without requiring 3D ground truth. Crucially, the inherent continuity of implicit representation overcomes sparse observations, enabling anatomically reliable geometry at arbitrary resolutions. Furthermore, to empower our framework with physically continuous 4D extension, we introduce a Radial SDF Alignment strategy that strictly locks shape evolution to the predicted velocity field, fundamentally eliminating mesh drift. Extensive experiments on synthetic benchmarks and real clinical datasets demonstrate that Echo4DIR achieves state-of-the-art 4D cardiac mesh reconstruction, notably yielding an impressive clinical overlap of up to 98.35% Dice and 96.75% IoU.

Distributed Image Compression (DIC) is crucial for multi-view transmission, especially when operating at extremely low bitrates (< 0.1 bpp). Its core challenge is effectively utilizing side information to achieve high-quality reconstruction under strict bitrate budgets. However, existing DIC approaches struggle to exploit global context and object-level details from side information, leading to local blurring and the loss of fine details in the reconstruction. To address these limitations, we propose a Multimodal DIC framework (MDIC), which, for the first time, leverages side information in a multimodal manner into the DIC paradigm, effectively preserving fine-grained local details and enhancing global perceptual quality in reconstructed images. Specifically, we introduce a text-to-image diffusion-based decoder conditioned on textual side information extracted from correlated images to capture shared global semantics. Moreover, we design a feature-mask generator, supervised by a multimodal fine-grained alignment task, to strengthen the exploitation of visual side information. The generated mask serves two purposes: first, it guides the extraction of fine-grained details from losslessly transmitted side information to preserve the semantic consistency of reconstructed details; second, it regulates the extraction of clustered feature representations from the quantized VQ-VAE embeddings, compensating for category information lost under the extreme compression of the primary image. Extensive experiments on the widely used KITTI Stereo and Cityscapes datasets demonstrate that MDIC achieves state-of-the-art perceptual quality at extremely low bitrates.

Generating high-fidelity visual effects (VFX) typically demands massive datasets and prohibitive computational power due to the intricate coupling of spatial textures and temporal dynamics. In this paper, we introduce EasyVFX, a resource-efficient framework that achieves realistic VFX synthesis under stringent constraints. Our core philosophy lies in frequency-domain decomposition: we observe that the complexity of VFX can be significantly mitigated by decoupling high-frequency components, which represent intricate spatial appearances, from low-frequency components that encapsulate global motion dynamics. This spectral disentanglement transforms a high-dimensional learning problem into manageable sub-tasks, thereby lowering the optimization barrier and reducing data dependency. Building upon this insight, we propose a two-stage training paradigm. First, we design a Frequency-aware Mixture-of-Experts (Freq-MoE) architecture. By utilizing a soft routing mechanism, our model assigns specialized experts to distinct spectral bands, enabling them to cultivate robust priors for appearance and motion dynamics. This specialization allows the model to acquire foundational VFX knowledge with fewer GPU resources. Second, we introduce a Test-Time Training strategy powered by a novel Frequency-constraint Loss. This allows the pre-trained model to swiftly adapt to specific, unseen effects through localized optimizations, requiring only about 100 steps on a single GPU. Experimental results demonstrate that EasyVFX produces structurally consistent and visually stunning effects, proving that frequency-aware learning is a key catalyst for democratizing professional-grade VFX.

Accurate localization of myocardial infarction is essential for risk stratification. While LGE-MRI remains the gold standard, it is resource-intensive. Integrating cine MRI with ECG enables a more detailed representation of infarct properties. Existing inverse MI inference methods overlook realistic scar morphology and cardiac repolarization, reducing sensitivity to subtle ECG variations and interpretability of infarct-induced electrophysiological changes. In this paper, we propose a novel framework for noninvasive MI localization using cardiac digital twins. To bridge the domain gap between simulation and reality, we introduce an anatomy-aware stochastic infarct synthesis strategy to synthesize realistic, irregular scars with border zones, mimicking ischemic transmural progression. We then construct a virtual cohort to simulate QRS-T waveforms, capturing both depolarization and repolarization dynamics. Furthermore, we design a Physiology and Anatomy Aware Network (PAA-Net) that jointly encodes 3D myocardial geometry and multi-lead ECGs to infer infarct areas with varying localizations, sizes, spatial extents, and transmuralities. Experimental results demonstrate that our framework significantly outperforms existing methods in inverse inference, achieving Dice scores of 0.7391 and 0.5503 for scar and border zone segmentation, respectively, while further enhancing the interpretability of the ECG-infarct relationship. Our code will be released upon acceptance.

Despite significant progress in Vision-Language Navigation (VLN), existing approaches still rely on dense RGB videos that produce excessive patch tokens and lack explicit spatial structure, resulting in substantial computational overhead and limited spatial reasoning. To address these issues, we introduce the Geometry-Aware BEV (GA-BEV) - a compact, 3D-grounded feature representation that integrates both explicit and implicit geometric cues into multimodal large language model (MLLM) - based navigation systems. We construct BEV spatial maps from RGB-D inputs by projecting visual features into 3D space and aggregating them into an agent-centric layout that preserves geometric consistency while reducing token redundancy. To further enrich geometric understanding, we incorporate features from a pretrained 3D foundation model into the BEV space, injecting structural priors learned from large-scale 3D reconstruction tasks. Together, these complementary cues - explicit depth-based projection and implicit learned priors - yield compact yet spatially expressive representations that substantially improve navigation efficiency and performance. Experiments show that our method achieves state-of-the-art results using only navigation data, without DAgger augmentation or mixed VQA training, demonstrating the robustness and data efficiency of the proposed GA-VLN framework.

Continual Visual Question Answering (VQA) requires learning from non-stationary streams of visual inputs and questions while preserving past knowledge. Most prior methods adapt by updating a largely shared parameter set. This often leads to cross-level task interference, hindering accurate adaptation to the current task and object. To address this limitation, we propose HyLoVQA. It maintains a drift-resilient memory bank of anchors. The bank stores the content of visual objects and textual tasks, and they are updated using current input features. Conditioned on retrieved anchors, a hypernetwork generates lightweight Low-Rank Adaptation (LoRA) adapters. This ensures parameter efficiency, allowing the model to adapt to each task and object dynamically. Additionally, we formulate an alignment loss that aligns semantic discrepancies in the feature space with functional changes in the parameter space, thereby constraining LoRA adapters to remain focused on the current task and object. Extensive experiments on VQA v2 and NExT-QA under both standard and compositional settings demonstrate the superiority of HyLoVQA over prior state-of-the-art methods.

Visual grounding, the task of localizing objects described by natural-language expressions, is a foundational capability for agricultural AI systems, enabling applications such as selective weeding, disease monitoring, and targeted harvesting. Reliable evaluation of agricultural visual grounding remains challenging because agricultural targets are often small, repetitive, occluded, or irregularly shaped, and instructions may refer to one, many, or no objects in an image. Evaluating this capability therefore requires jointly testing localization accuracy, target-set completeness, and existence-aware abstention. To address these challenges, we introduce \textbf{AgroVG}, a multi-source benchmark that formulates agricultural grounding as generalized set prediction: given an image and a referring expression, a model must return all matching target instances or abstain when no target is present. AgroVG contains 10{,}071 annotation-grounded image-query pairs from ten source datasets across six target families: crop/weed, fruit, wheat head, pest, plant disease, and tree canopy. It supports bounding-box grounding (T1) across all six families and instance-mask grounding (T2) on sources with reliable instance-level pixel annotations, with queries covering single-target, multi-target, and target-absent regimes. AgroVG further provides task-specific protocols for box-set matching and query-level mask coverage. Zero-shot evaluation of 26 model configurations spanning closed-source MLLMs, open-source VLMs, and specialized grounding systems reveals persistent gaps: the best multi-target Set-$F_1$ reaches only 0.35, and the best positive-query mask success rate at IoU@0.75 remains below 0.17. Data and code are available at https://anonymous.4open.science/r/AgroVG-5172/ .

Accelerated MRI reconstruction requires recovering missing details while preserving anatomically coherent structures across large spatial regions. State-space models such as Mamba provide efficient long-range modeling, making them attractive learned regularizers for unrolled reconstruction. However, in a data-consistency-coupled unrolled solver, different stages operate on different reconstruction iterates, where the resident carrier should preserve coherent reconstruction content across stages while stage-dependent non-resident evidence is tied to the current update. Treating these roles uniformly can place persistent resident-carrier evidence and update-dependent non-resident evidence into the same recurrent content route. We therefore propose SO-Mamba, a state-ownership Mamba regularizer that assigns reconstruction evidence within each Mamba stage to recurrent residency, state-interface access, and non-state output correction. SO-Mamba implements this ownership rule with a State-Ownership Router (SOR), which constructs a resident carrier for recurrent content and routes non-resident evidence to affine modulation of the B/C state interfaces and an output correction outlet. The resident carrier supplies the Mamba content route, while the non-resident evidence stream adapts the state interfaces and contributes through the output outlet without entering the recurrent content route. We further introduce a two-level outer-band leakage diagnostic that separates hidden-state storage from readout expression by measuring outer-band energy in the selective-scan state trajectory and the post-scan Mamba readout. Experiments on five public MRI reconstruction benchmarks spanning diverse anatomies, sampling patterns, and coil configurations show that SO-Mamba consistently improves over CNN-, Transformer-, and Mamba-based baselines with competitive computational efficiency.

Deepgenerative models havebecomeapromisingapproach for human motion prediction due to their ability to capture multimodal distributions and represent diverse human be haviors. However, generating predictions that are both di verse and jointly consistent among interacting agents re mains challenging. In addition, most existing approaches are primarily evaluated using single-agent (marginal) met rics, which fail to fully reflect the joint dynamics of multi agent interactions. We propose a diffusion-based frame work that improves multi-agent motion prediction by lever aging rich contextual information from historical trajecto ries. This information is incorporated through a guidance mechanism to enhance the diversity and expressiveness of predicted motions. To further enforce interaction consis tency, we introduce an energy-based formulation that re fines the joint trajectory distribution while preserving the plausibility of individual trajectories. Extensive experi ments on four benchmark datasets demonstrate that our approach consistently outperforms existing methods. No tably, our approach substantially improves both marginal (ADE/FDE) and joint (JADE/JFDE) metrics on ETH/UCY over strong marginal baselines. Compared with prior joint prediction methods, it delivers significant gains in marginal metrics while maintaining competitive joint performance.

Diffusion Transformer (DiT) has emerged as a powerful model architecture for generating high-quality images and videos. In the case of video DiT, 3D Spatio-Temporal Attention increases token length in proportion to the number of frames, sharply increasing computational cost. Token reduction methods mitigate this cost by exploiting spatial redundancy, but existing approaches rely on inaccurate similarity estimates and lightweight matching algorithms, resulting in poor matching quality and only marginal acceleration. To overcome these limitations, we propose ORBIS, an SW-HW co-designed accelerator for video DiT. ORBIS leverages the output activation from the previous timestep to obtain more accurate inter-token similarity, substantially improving matching quality and enabling a higher token reduction ratio. We further introduce a Distribution-Aware Token Matching (DATM) algorithm that captures global token distribution and explicitly minimizes token-pair loss for additional gains. To fully hide DATM latency, we design specialized, deeply pipelined hardware and minimize its hardware cost through quantization, occupying only 2.4% of total area with negligible accuracy loss. Extensive experiments show that ORBIS achieves about 2x higher token reduction ratio than the state-of-the-art approach, AsymRnR, while delivering up to 4.5x speedup and 79.3% energy reduction compared to an NVIDIA A100 GPU.

Understanding 3D point clouds through language remains a fundamental challenge in computer graphics and visual computing, due to the irregular structure of point cloud data and the lack of explicit reasoning in existing 3D multimodal models. While Chain-of-Thought (CoT) reasoning has shown strong effectiveness in LLMs and image-based MLLMs, its extension to 3D understanding remains largely underexplored. In this paper, we propose a data-centric framework for constructing large-scale CoT supervision tailored to 3D point cloud understanding. Our framework consists of a two-stage pipeline that first refines point-text instruction data via vision-language-model-based quality evaluation and reference-guided refinement, and then synthesizes high-quality reasoning paths through Human-in-the-Loop Prompt Optimization (HiLPO). Using this approach, we build PoCoTI, a CoT-enhanced point-text instruction-following dataset containing 55K samples with explicit reasoning paths. Fine-tuning PointLLM on PoCoTI yields PointLLM-R, a reasoning-capable 3D multimodal language model. Extensive experiments on generative 3D classification and captioning demonstrate that PointLLM-R achieves state-of-the-art performance and generalizes robustly to real-world scanned point clouds and multi-turn dialogue scenarios.

Joint audio-visual reasoning is essential for omnimodal understanding, yet current multimodal large language models (MLLMs) still struggle when reasoning requires fine-grained evidence from both modalities. A central limitation is that explicit text-based chain-of-thought (CoT) compresses continuous audio-visual signals into discrete tokens, weakening temporal grounding and shifting intermediate reasoning toward language priors. We argue that a unified latent space is a better medium for such reasoning because it preserves dense sensory information while remaining compatible with autoregressive generation. Based on this insight, we propose \textbf{LatentOmni}, a cross-modal reasoning framework that interleaves textual reasoning with audio-visual latent states. LatentOmni introduces feature-level supervision to align latent reasoning states with task-relevant sensory features and uses Omni-Sync Position Embedding (OSPE) to maintain temporal consistency between latent audio and visual states. We further construct \textbf{LatentOmni-Instruct-35K}, a dataset of audio-visual interleaved reasoning trajectories for supervising latent-space reasoning. Comprehensive evaluation across multiple audio-visual reasoning benchmarks demonstrates that LatentOmni achieves the best performance among the evaluated open-source models and consistently outperforms the Explicit Text CoT baseline, supporting latent-space joint reasoning as a promising path toward stronger omnimodal understanding.

Diffusion Transformers (DiTs) achieve superior image generation quality but suffer from quadratic computational complexity relative to token count. While various token reduction (TR) methods have been proposed to mitigate this cost, they overlook the primary objective of generative models: minimizing recovery error, which requires reflecting output token similarity. They rely solely on input token similarity inherited from reduction-only ViT paradigms, leading to a fundamental misalignment with this objective. To bridge this gap, we propose DiTo, a novel TR paradigm that shifts the focus toward output-centric token reduction. Based on the observation that output token similarity is consistently preserved across adjacent timesteps, DiTo utilizes prior-step similarities as an effective proxy to establish token correspondences at a Matching timestep, which are then reused across multiple subsequent Reduction timesteps. To optimize this interleaved scheduling, we propose Pair Match Ratio (PMR)-guided Interval Scheduling to determine the optimal matching frequency. Furthermore, to mitigate localized approximation errors and resulting blocking artifacts caused by repeated reuse, we propose Frequency-aware Token Matching by incorporating a selection-frequency penalty. Extensive experiments demonstrate that DiTo consistently outperforms existing TR methods with 1.6-3.9 dB higher PSNR at comparable speedups, achieving a superior Pareto frontier.

Biomedical image segmentation is a critical task in medical diagnosis and treatment planning, enabling precise delineation of anatomical structures and pathological regions. Despite significant advancements, challenges persist due to the inherent variability, noise, and complex morphology present in diverse medical imaging modalities. This paper introduces ConvNeXt-FD, a novel deep learning architecture for robust biomedical image segmentation, built upon a U-Net-like encoder-decoder framework leveraging the powerful ConvNeXt backbone. Our approach integrates a hybrid loss function combining the Dice coefficient with a boundary-aware regularization term inspired by a differentiable formulation of Fractal Dimension, designed to enhance the model's sensitivity to object boundaries and shape fidelity. We rigorously evaluate ConvNeXt-FD across six distinct biomedical datasets: BUSI (Breast Ultrasound Images), DDTI (Thyroid Ultrasound Images), FluoCells (Fluorescent Cell Images), IDRiD (Diabetic Retinopathy Images for Optic Disc Segmentation), ISIC2018 (Skin Lesion Images), and MoNuSeg (Nuclei Segmentation). Experimental results demonstrate that ConvNeXt-FD, particularly when initialized with ImageNet pre-trained weights, achieves competitive and often superior performance compared to existing state-of-the-art methods across various metrics, including Dice, Jaccard, Accuracy, Sensitivity, Specificity, and False Positive Rate. The integration of ConvNeXt as a strong encoder, coupled with the boundary-aware regularization, proves effective in capturing both high-level semantic features and fine-grained boundary details, leading to more accurate and reliable segmentations in challenging biomedical contexts.

Three-dimensional (3D) histopathology of unprocessed tissues has the potential to transform disease management by enabling volumetric characterization of tissue microarchitecture and in-vivo assessment. Back-illumination Interference Tomography (BIT) is a new phase microscopy technology that provides rapid, non-destructive volumetric imaging of unprocessed tissues. However, translating BIT volumes into clinically interpretable H&E images remains challenging, particularly due to shift-variant contrast and the absence of quantitative validation benchmarks. We introduce HistoBIT3D, the first voxel-wise paired BIT and fluorescence-labeled nuclei dataset, enabling quantitative evaluation of structural preservation in unsupervised virtual staining against ground-truth nuclear distributions. Using this dataset, we present a novel virtual staining framework that translates BIT volumes with shift-variant contrast into realistic H&E volumes by leveraging bidirectional multiscale content consistency and cross-domain style reuse to enhance structural fidelity and perceptual realism. Our method achieves state-of-the-art realism metrics while significantly improving 3D nuclei segmentation accuracy and boundary preservation under zero-shot Cellpose evaluation. Together, these contributions establish a quantitatively validated, structurally faithful, and scalable pipeline for 3D virtual H&E staining, advancing the paradigm of slide-free, volumetric computational histopathology. Our data and code are available at: https://github.com/aasong113/HistoBIT3D_VirtualStaining.

Video large language models (Video LLMs) achieve strong benchmark accuracy, yet often answer video questions through shortcuts such as single-frame cues and language priors rather than by tracking spatiotemporal dynamics. This issue is exacerbated in RL post-training, where correctness-only rewards can further reinforce shortcut policies that obtain high reward without tracking video dynamics. We address this by asking a controlled counterfactual question: if the visual world changed while the question remained fixed, should the answer change or stay the same? Based on this view, we propose \textbf{Counterfactual Relational Policy Optimization (CRPO)}, a dual-branch RL framework for improving \emph{spatiotemporal sensitivity}. CRPO constructs counterfactual videos through horizontal flips and temporal reversals, trains on both original and counterfactual branches, and introduces a \textbf{Counterfactual Relation Reward (CRR)} between their answers. CRR encourages answers to change for dynamic questions and remain unchanged for static questions. This cross-branch constraint makes it difficult for shortcut policies to be consistently rewarded across both branches. To evaluate this property, we introduce \textbf{DyBench}, a paired counterfactual video benchmark with 3,014 videos covering reversible dynamics, moving direction, and event sequence, together with a strict pair-accuracy metric that prevents fixed-answer shortcuts from inflating scores. Experiments show that CRPO outperforms prior RL methods on spatiotemporal-sensitive evaluations while maintaining competitive general video performance. On Qwen3-VL-8B, CRPO improves DyBench P-Acc by +7.7 and TimeBlind I-Acc by +8.2 over the base model, indicating improved spatiotemporal sensitivity rather than stronger reliance on static shortcuts. The project website can be found at https://ddz16.github.io/crpo.github.io/ .

Flow matching with $x$-prediction -- regressing the clean data point rather than the ambient velocity -- is known to exploit low-dimensional manifold structure effectively in pixel space \cite{li2025back}. We ask whether a pretrained representation space, while containing a low-dimensional data manifold of comparable intrinsic dimensionality, offers a distribution more favorable for flow-matching learning. Comparing pixel, SD-VAE, and DINOv2 features along four geometric axes, we find that pixel and DINOv2 share nearly identical intrinsic dimensionalities (both $\hat{d}\!\approx\!33$) yet DINOv2 exhibits $7.3\times$ higher effective rank, $35\times$ better covariance conditioning, $11.5\times$ lower excess kurtosis, and $1.7\times$ lower on-manifold interpolation error; SD-VAE latents are consistently intermediate, indicating that the advantage stems from representation-learning objectives rather than mere compression. These statistical properties render the flow-matching regression well-conditioned and remove the need for the specialized prediction heads or Riemannian transport used by prior DINOv2 diffusion methods. We propose the \emph{Representation Image Transformer} (RiT): a vanilla Diffusion Transformer trained by $x$-prediction on frozen DINOv2 features, augmented only by a dimension-aware noise schedule and joint \texttt{[CLS]}-patch modeling. On ImageNet $256{\times}256$, RiT attains FID 1.45 without guidance and 1.14 with classifier-free guidance, outperforming DiT$^\text{DH}$-XL with $19\%$ fewer parameters (676M vs.\ 839M). The resulting ODE is efficiently solvable at coarse discretizations: with classifier-free guidance, $5$ Heun steps already reach FID 2.0 and $10$ steps reach 1.25, without distillation or consistency training. Code at https://github.com/lezhang7/RiT.

Large Vision-Language Models (LVLMs) represent a significant leap towards empathetic agents, demonstrating remarkable capabilities in emotion understanding. However, the internal mechanisms governing how LVLMs translate abstract visual stimuli into coherent emotional narratives remain largely unexplored, primarily due to the scarcity of visual counterfactuals and the diffuse nature of emotional expression. In this paper, we bridge this gap by introducing a steering-vector-based causal attribution framework tailored for descriptive emotional reasoning. To this end, we construct a specialized dataset to demystify the emotional circuits underlying the three-stage ``Adapt-Aggregate-Execute'' mechanism. Crucially, we discover a functional decoupling: visual emotional cues are aggregated in middle layers via sentiment-specific attention heads, but are subsequently translated into narrative generation in deep layers through emotion-general pathways. Guided by these insights, we regulate the emotional information routing to strengthen attention flow and amplify the semantic activation to consolidate expression. Extensive experiments on the comprehensive MER-UniBench demonstrate that our methods significantly improve performance via inference-time intervention, effectively mitigating emotional hallucinations and corroborating the causal fidelity of the discovered circuits.

AI-generated content (AIGC) is rapidly improving, creating an urgent need for detectors that generalize across data sources, deployment pipelines, and visual modalities. A strongly generalizable detector should remain robust under distributional variations. However, we identify a consistent failure mode: SOTA AI-generated image detectors often collapse when applied to frames extracted from videos. Through systematic analysis, we show that this cross-modal gap arises from both entangled synthesis-agnostic video processing shifts, including color conversion, codec compression, resizing, and blur, and model-specific fingerprints introduced by modern video generators. Motivated by these findings, we propose VINA (Video as Natural Augmentation), a unified AIGC detection framework that jointly trains on image and video data. VINA uses video frames as physically grounded natural augmentations and further introduces a cross-modal supervised contrastive objective to align image and video representations under a shared real/fake decision boundary. Extensive experiments on 14 image, video, and in-the-wild benchmarks show that VINA delivers bidirectional gains, improves robustness and transferability, and achieves state-of-the-art performance across nearly all evaluated settings without complex augmentation or dataset-specific tuning.

Current Video-LLM approaches for Video Temporal Grounding (VTG) typically rely on direct timestamp generation from an unstructured visual-token stream, often leading to brittle numerics and inconsistent boundaries. To address this, we propose Foresee-to-Ground (F2G), a framework that reformulates VTG as a verifiable Identify-then-Measure problem. F2G integrates Predictive Temporal Perception with Evidence-Driven Reasoning: it learns boundary-sensitive temporal representations to build a video-wide evidence pool of candidate event segments, and exposes these segments to the LLM as citable evidence units that bind boundary prediction to explicit event hypotheses. By decoupling event identification from precise boundary measurement, F2G stabilizes grounding and makes predictions verifiable. Extensive experiments demonstrate that F2G consistently improves grounding accuracy across diverse benchmarks, transfers robustly across different Video-LLM backbones, and preserves general video understanding capabilities.

Accurate and robust medical image classification is paramount for early disease diagnosis and treatment planning. However, challenges such as limited annotated data, high intra-class variability, and subtle inter-class differences often hinder the performance of deep learning models. This paper introduces a synergistic deep learning framework that leverages the strengths of self-supervised learning and transfer learning for enhanced medical image classification. Our approach employs two distinct ConvNeXt-Tiny models: one pre-trained on a large-scale natural image dataset (ImageNet) and another pre-trained using an entropy-guided Masked Autoencoder (MAE) on the target medical dataset. Both models are then fine-tuned on specific medical image classification tasks. A final ensemble strategy, based on averaging predicted probabilities, is utilized to combine the complementary insights from these two models. Rigorous experimental validation across four diverse medical imaging datasets (Breast Ultrasound Images (BUSI), International Skin Imaging Collaboration (ISIC) 2018, Kvasir, and COVID) demonstrates the superior performance and robustness of our ensemble approach. The MAE pre-training significantly improves feature learning on domain-specific data, while the ImageNet pre-training provides strong generalizable features. The ensemble consistently achieves state-of-the-art results, outperforming individual models and existing methods, highlighting the efficacy of combining diverse pre-training strategies for challenging medical image analysis.

Video anomaly detection is critical for public safety and security, yet remains highly challenging despite extensive research due to large variations in appearance, viewpoint, and scene dynamics. Among existing approaches, human pose-based methods have emerged as a major line of research, showing strong performance since many anomalies in public datasets involve humans and pose representations are robust to appearance changes while providing compact motion descriptions. However, these methods often overlook bounding-box trajectories, although such information is inherently available in pose-based pipelines. In this paper, we explicitly leverage these trajectories as a primary anomaly cue. We present TrajVAD, a framework that models multi-class bounding-box trajectories using normalizing flows to learn normal kinematic patterns. Its trajectory-only variant (TrajVAD-T) eliminates pose estimation and surpasses all compared pose-based methods on ShanghaiTech in AP (87.7%), while achieving the best results on MSAD. An extended version (TrajVAD-P) incorporates pose information and further improves performance to 88.6% AUROC and 90.9% AP on ShanghaiTech, highlighting bounding-box trajectories as an effective yet underexplored modality for video anomaly detection.

Video temporal grounding (VTG), which localizes the start and end times of a queried event in an untrimmed video, is a key test of whether multimodal large language models (MLLMs) understand not only what happens but also when it happens. Although modern MLLMs describe video content fluently, their timestamp predictions remain unreliable, while existing remedies either require costly post-training on temporal annotations or rely on coarse training-free heuristics. In this work, we probe the cross-modal attention of MLLMs and uncover a perception-generation gap. Our key finding is that MLLMs often know the target interval during prefill, but lose this signal when generating the final answer. In the prefill stage, a sparse set of attention heads, which we call \emph{Temporal Grounding Heads} (TG-Heads), concentrates query-to-video attention on the ground-truth interval. During autoregressive decoding, however, the answer tokens shift attention away from this interval toward visually salient but query-irrelevant segments. This observation motivates an inference-time read-then-regenerate framework. We first convert TG-Head prefill attention into a debiased frame-level relevance signal and extract the high-attention interval it highlights. We then re-invoke the MLLM with visual context restricted to this interval, using video cropping or attention masking to suppress distractors. Without parameter updates and architectural changes, our framework consistently improves MiMo-VL-7B, Qwen3-VL-8B, and TimeLens-8B on three VTG benchmarks, with gains of up to +3.5 mIoU. The project website can be found at https://ddz16.github.io/mllmsknowwhen.github.io/.

Recent Video Large Language Models (Video-LLMs) have demonstrated strong capabilities in video reasoning through reinforcement learning (RL). However, existing RL pipelines rely heavily on human-annotated tasks and solutions, making them costly to scale and fundamentally constrained by human expertise. Self-evolving frameworks have recently emerged as a promising alternative through autonomous Questioner-Solver self-play. Unfortunately, these approaches are primarily designed for static modalities such as text and images, fundamentally failing to capture the temporal dynamics that are central to video reasoning. In this work, we propose $\textbf{EvoVid}$, a temporal-centric self-evolving framework that enables Video-LLMs to improve directly from raw, unannotated videos. Specifically, we introduce two complementary temporal-centric rewards: a temporal-aware Questioner reward that encourages temporally dependent question generation through temporal perturbation sensitivity, and a temporal-grounded Solver reward that provides automatic temporal supervision via inherent video segment localization. Extensive experiments across four base models and six benchmarks demonstrate consistent improvements over both base models and existing self-evolving baselines, achieving competitive performance with supervised methods. These results highlight temporal-centric self-evolution as an effective and scalable paradigm for video understanding and reasoning.

On-policy knowledge distillation has proven effective for language models, yet its application to vision-language models (VLMs) remains underexplored. We observe that standard on-policy distillation can improve a student's output quality while failing to strengthen its reliance on visual input: on vision-critical tokens, the student's predictions remain largely unchanged whether or not fine-grained visual detail is present, even though the teacher's predictions depend heavily on it.To make this difference observable, we introduce visual advantage (VA), the token-level log-probability difference when the teacher scores a student-generated rollout with versus without access to fine-grained visual detail. VA is concentrated in a small minority of tokens, and these high-VA tokens are the ones that actually carry the visual supervision signal. This motivates a distillation objective that treats them differently from language scaffolding, so their contribution is not diluted by the abundant surrounding language tokens.We propose Visual-Advantage On-Policy Distillation (VA-OPD), which uses VA at two granularities: rollout-level reweighting by trajectory-averaged VA, and token-level KL averaged within high-VA and low-VA groups separately. We train on two math datasets (Geometry3K and ViRL39K) and evaluate on eight benchmarks covering both mathematical reasoning and visual understanding, across three teacher sizes (4B, 8B, and 32B) on the Qwen3-VL family. VA-OPD improves over standard on-policy distillation on every benchmark, with the gain growing monotonically along both the teacher-size and data-scale axes, suggesting that these factors compound consistently.

Assessing progress toward the Sustainable Development Goals (SDGs) requires multi-step reasoning over visual cues, contextual knowledge, and development indicators, where incomplete evidence use and imperfect evidence integration can introduce hidden prediction biases. Real-world SDG monitoring further spans both qualitative judgments and quantitative estimation. However, existing benchmarks typically evaluate these aspects in isolation, obscuring systematic biases that emerge when models substitute priors for evidence. To address this gap, we propose SDGBiasBench, a large-scale benchmark suite for SDG-oriented vision-language reasoning. Spanning 500k expert-involved multiple-choice questions and 50k regression tasks, the benchmark enables comprehensive assessment of both decision-level and estimation-level bias in Vision--Language Models (VLMs). Evaluations on SDGBiasBench reveal an intrinsic SDG bias in current VLMs, where predictions are frequently driven by SDG specific priors rather than reliable multi-modal cues. To mitigate such bias, we propose CADE (Contrastive Adaptive Debias Ensemble), a training-free, plug-and-play method that leverages modality-specific answer priors. CADE yields significant gains on the proposed benchmark, improving multiple-choice accuracy by up to 25% and reducing regression MAE by up to 12 points across multiple VLMs. We hope our work can foster the development of more fair and reliable AI systems for sustainable development.

Training Vision Language Models (VLMs) for video event reasoning requires high-quality structured annotations capturing not only what happened, but when, where, why, and with what consequence, at a scale manual labelling cannot support. We present MAVEN (Multi-stage Agentic Video Event aNnotation), a multi-stage agentic pipeline that turns raw videos into multi-task training data with Chain-of-Thought (CoT) reasoning traces, organized around a designated Event of Focus. At its core, MAVEN synthesizes a Multi-Scale Spatio-Temporal Event Description (MSTED) from three complementary caption levels; this explicit intermediate serves as the sole input to downstream Q&A generation across multiple task formats. Crucially, MAVEN supports agent-driven domain adaptation: given a new video dataset and target question examples, the agent redesigns all prompts top-down without manual re-engineering. A hierarchical refinement loop further classifies annotation errors against a taxonomy, traces root causes to the originating pipeline stage, and applies targeted edits that rewrite prompts or modify the pipeline structure itself, iteratively improving data quality. We apply MAVEN to label over 5,300 traffic videos and fine-tune Cosmos-Reason2-8B on the resulting data. On a private CCTV evaluation set, fine-tuning surpasses both Gemini 2.5 Pro and 3.1 Flash, including a $+38.8$-point gain in MCQ accuracy over zero-shot. On AccidentBench, CCTV-only training lifts Cosmos-Reason2 by $+10.7$ MCQ points and matches Gemini 2.5 Pro despite seeing no dashcam videos; adding agent-adapted dashcam annotations narrows the gap to Gemini 3.1 Flash, and RL post-training pushes overall performance past both Gemini baselines. Qualitative results on warehouse surveillance and public safety videos further show the agentic workflow readily adapts the pipeline to new domains.

Stereo image super-resolution aims to generate high-resolution images by leveraging complementary information from binocular systems. Although previous studies have achieved impressive results, the potential of intra-view and cross-view information has not been fully exploited. To address this issue, we propose a novel multi-scale interaction network for stereo image super-resolution. Specifically, we design a Multi-scale Spatial-Channel Attention Module that utilizes multi-scale large separable kernel attention and simple channel attention to improve intra-view feature extraction. Additionally, we propose a Dual-View Epipolar Attention Module, utilizing an optimal transport algorithm to achieve more accurate matching along the epipolar line. Extensive experimental and ablation studies show that our method achieves competitive results that outperform most SOTA methods.

The efficient Test-Time Scaling (TTS) paradigm offers a promising perspective for enhancing the generation performance of diffusion models. However, current solutions are limited to a static, pre-defined noise pool and suffer from inflexible noise exploration across the denoising trajectory. To bridge this gap, we propose RTS, a novel Reward-guided Trajectory Scaling method to fully unlock the generative potential of diffusion models. Unlike existing methods, RTS facilitates the synthesis of refined, high-fidelity images via two core innovations: 1) a reward-guided noise optimization strategy to actively direct the search towards promising regions; and 2) a sparse test-time scaling framework together with a PCA-driven curvature analysis scheme to prioritize key intermediate steps in the entire denoising space, effectively compressing the search space. Experiments show our approach outperforms baselines by 15.6% across GenEval Score, and a 60.4% enhancement in ImageReward score, setting a new SOTA while providing a practical guideline for more effective test-time scaling across diffusion-specific architectures.

Vision-language models (VLMs) often fail under low illumination because their visual grounding is learned predominantly from RGB imagery, whereas thermal infrared preserves complementary scene structure when visible cues degrade. We present Thermo-VL, a wavelength-aware VLM that augments a frozen Molmo-7B backbone with a trainable thermal encoder and a text-guided dual-attention fusion module. Given aligned RGB tokens, thermal tokens, and prompt embeddings, the fusion module conditions thermal features on both language and RGB context, then injects a gated residual into the frozen RGB stream so thermal evidence can be incorporated without disrupting Molmo's pretrained RGB-language interface. We train the model with the standard language-modeling objective together with auxiliary alignment and regularization losses that improve cross-modal grounding and reduce over-reliance on RGB. We also introduce a pixel-aligned RGB-thermal instruction-tuning dataset and Thermo-VL-Bench, a manually screened RGB-thermal VQA benchmark for low-light and cross-spectrum reasoning. Experiments show strong gains on challenging thermal-only and RGB+thermal reasoning tasks, highlighting the value of prompt-conditioned multispectral fusion. Our dataset and code are publicly available at: https://thusharakart.github.io/Thermo-VL

We present our submission to the Hume-ABAW10 Emotional Mimicry Intensity (EMI) Challenge, which aims to predict six continuous emotion intensity dimensions: Admiration, Amusement, Determination, Empathic Pain, Excitement, and Joy, from in-the-wild multimodal video clips. We propose a staged multimodal framework that combines textual, acoustic, and visual representations, with an optional motion branch. Our approach first trains modality-specific encoders independently and then fuses their learned representations through a lightweight regressor with modality dropout and controlled encoder adaptation. Across our submitted systems, the best validation performance is obtained by the text--audio--vision--motion fusion model under the expanded 4:1 split, achieving an average Pearson correlation of 0.4722. Although the motion branch yields only very slight gains, its behavior can be interesting to study. Our team was placed third in the EMI challenge, achieving an average Pearson correlation of 0.57 for the test set. Overall, we provide a practical and reproducible baseline for EMI prediction.

Multi-modality medical vision (MV) foundation models (FM) are fundamentally challenged by pronounced Non-IID feature statistics across heterogeneous imaging modalities. Monolithic self-supervised optimization on such data induces conflicting gradients, driving representations to collapse toward modality-dominant shortcuts. This work reframes this failure as an imbalance between specialization and coordination in emergent modularity, and proposes Director-Experts (DEX), a modular network that explicitly regulates these dynamics in stacked modules. Each DEX module comprises a pool of experts, dynamically adapted by our image-wise activation strategy, autonomously specializing in modality-dominant statistics, together with a director, updated via our group exponential moving average, which distills multi-expert knowledge into a shared space for semantic integration across modalities, thus driving the emergence of modular representations. We curate a new benchmark, Medical Vision Universe, over 4 million images across 10 modalities, which provides a FM-level pre-training with the broadest coverage of distinct imaging modalities to our DEX. Extensive evaluations on 26 downstream tasks demonstrate improved optimization behavior and transferability, indicating DEX as a principled step toward general-purpose multi-modality medical AI. Our code and dataset will be opened at https://github.com/YutingHe-list/DEX.

While Multimodal Large Language Models (MLLMs) have demonstrated remarkable proficiency in general video understanding, their capacity to interpret involuntary, and spatio-temporally evolving pathologic motor behaviors such as seizure semiology remains largely untested. To address this gap, we introduce Seizure-Semiology-Suite, a clinically grounded dataset and benchmark for fine-grained, structured seizure semiology understanding. The dataset includes 438 seizure videos annotated with over 35,000 dense labels covering 20 ILAE-defined semiological features. Building on this dataset, we propose a seven-task hierarchical benchmark that systematically evaluates MLLMs from low-level visual perception to temporal sequencing, narrative report generation, and seizure diagnosis. To enable clinically meaningful evaluation of generated reports, we further introduce the Report Quality Index for Seizure Semiology (Seizure-RQI). Extensive baselines across 11 open-weight MLLMs reveal systematic weaknesses in laterality reasoning, temporal localization, symptom sequencing, and clinically faithful reporting. We show that seizure-specific fine-tuning substantially improves performance across tasks, and that a two-stage neuro-symbolic framework achieves an F1 score of 0.96 on epileptic versus non-epileptic seizure classification. Seizure-Semiology-Suite establishes a rigorous benchmark for evaluating multimodal models in safety-critical medical video understanding and guides the development of clinically reliable, domain-adaptive multimodal intelligence.

The synergistic interpretation of anatomical information from computed tomography (CT) and metabolic information from positron emission tomography (PET) is important to oncologic imaging. However, existing deep learning methods for PET/CT remain largely task-specific, are often trained on single-center cohorts, or adopt dual-branch fusion schemes that delay cross-modal interaction and underutilize early spatial correspondence between PET and CT. To address these limitations, we present an open-source, multi-center, whole-body FDG PET/CT foundation model utilizing 4,997 harmonized scans from four public datasets. Our framework employs hierarchical UNet-shaped backbones with early channel-wise concatenation, enabling anatomical and metabolic features to interact from the first embedding layer onward. We further introduce a masked autoencoding objective based on zero-mean imputation, combined with a weighted global reconstruction loss. This design avoids non-physical intensity discontinuities at masked-region boundaries that arise from learnable mask tokens. On downstream AutoPET lesion segmentation, the proposed models demonstrate strong label efficiency: with only 10\% of the labeled training data, they achieve performance comparable to models trained from scratch on the full dataset. Under extreme 5-shot linear probing, joint PET/CT pretraining also achieves higher Dice scores than separated-modality pretraining. This multi-center foundation model demonstrates label efficiency and cross-modality representation learning for PET/CT tumor segmentation. It provides a robust, open-source basis for advancing automated oncologic imaging, significantly reducing the need for large-scale manual annotations in clinical practice.

Field-scale crop maps support supply-chain forecasting and policy, yet statewide crop identification still often depends on retrospective surveys or remote-sensing workflows built around hand-engineered spectral features. Those pipelines can be accurate, but they require repeated preprocessing and often lose robustness across years. This study evaluated whether Google DeepMind's AlphaEarth geospatial embeddings can serve as an analysis-ready alternative for mapping processing tomato systems in California. LandIQ 2018 crop polygons were used to assemble a balanced reference dataset of 4,742 tomato and 4,742 non-tomato fields. For each polygon, 64-band AlphaEarth embedding chips were extracted and aligned with binary masks, then divided into spatially independent training (n = 6,638), validation (n = 1,422), and test (n = 1,424) sets. A U-Net segmentation model was trained on AWS SageMaker using a composite masked binary cross-entropy and soft Dice loss. To complement hard predictions, Monte Carlo dropout was retained at inference and repeated 100 times per chip to estimate predictive mean and variance. On the independent test set, the model achieved 99.19% pixel accuracy, 98.69% precision, 99.40% recall, 99.04% F1 score, 98.11% intersection over union, and 99.02% chip accuracy. Uncertainty maps were consistently highest near field edges and low within field interiors. The results show that AlphaEarth embeddings retain crop-relevant spatial and temporal structure and can support accurate, field-scale tomato mapping without manual feature engineering.

Grounding language in the physical world requires AI systems to interpret references that emerge dynamically during conversation. While current vision-language models (VLMs) excel at static image tasks, they struggle to resolve ambiguous expressions in spontaneous, multi-turn dialogue. We address this gap by introducing (1) a benchmark for referential communication in dynamic 3D environments, built from 6.7 hours of egocentric VR interaction with synchronized speech, motion, gaze, and 3D scene geometry, and (2) a two-stage grounding pipeline that explicitly resolves conversational ambiguity before visual localization. The benchmark includes over 4,200 manually verified referring expressions spanning full, partitive, and pronominal types. Our contextual rewriting approach improves grounding performance by 11-22 percentage points on average, with a pure detector (GroundingDINO) reaching 56.7% on pronominals after rewriting, nearly double the best end-to-end baseline. Results demonstrate that decoupling linguistic reasoning from visual perception is more effective than end-to-end approaches for conversational grounding.

Zero-shot 3D visual grounding requires localizing objects in unstructured environments from free-form natural language. Recent vision-language model (VLM) approaches achieve promising results but rely on view-dependent reasoning or implicit representations, limiting spatial consistency and interpretability for compositional queries. We propose SceneGraphGrounder, a framework that reformulates 3D grounding as structured graph matching over a reconstructed 3D scene graph. To enable this formulation, we introduce a visual marker prompting strategy that enables a VLM to infer object-object relationships from 2D views, which are subsequently lifted into a persistent 3D scene graph encoding both spatial and semantic relations. Given a query, we construct a query graph and perform constrained alignment with the scene graph, ensuring multi-view consistency and interpretable reasoning. Experiments on the ScanRefer benchmark demonstrate that our method achieves competitive performance among zero-shot approaches, using only RGB-D inputs. We further validate our framework through real-world deployment on a mobile robot, demonstrating robust spatial reasoning in long-horizon physical environments. We will make our code publicly available upon acceptance.

Being able to relight human performance is a fundamental task for post production and content creation. We present BodyReLux, a subject-specific video diffusion-based framework for relighting full-body human performances in a temporally consistent way. Our model is trained on a hybrid dataset of pixel-aligned video relighting pairs, covering a diverse combination of lighting conditions, performances and viewpoints. To acquire such dataset, we combine traditional static One-Light-at-a-Time (OLAT) capture and a novel dynamic performance capture in which two smoothly varying lighting sequences are rapidly interleaved. Because the lighting operates above the human flicker-fusion threshold, the interleaving does not appear to strobe. We train our video relighting model from a pretrained text-to-video model to fully leverage the generative priors for producing high quality videos. To achieve accurate lighting control, we introduce a new lighting conditioning method that represents each light source as a token. We further condition on sequences of lighting using masked attention to support dynamic lighting control. Together with a carefully designed data augmentation pipeline, we achieve photorealistic, robust, and temporally consistent video relighting of subject-specific human performances.