Text-to-image (T2I) diffusion models achieve impressive photorealism by training on large-scale web data, but models inherit cultural biases and fail to depict underrepresented regions faithfully. Existing cultural benchmarks focus mainly on object-centric categories (e.g., food, attire, and architecture), overlooking the social and daily activities that more clearly reflect cultural norms. Few metrics exist for measuring cultural faithfulness. We introduce CULTIVate, a benchmark for evaluating T2I models on cross-cultural activities (e.g., greetings, dining, games, traditional dances, and cultural celebrations). CULTIVate spans 16 countries with 576 prompts and more than 19,000 images, and provides an explainable descriptor-based evaluation framework across multiple cultural dimensions, including background, attire, objects, and interactions. We propose four metrics to measure cultural alignment, hallucination, exaggerated elements, and diversity. Our findings reveal systematic disparities: models perform better for global north countries than for the global south, with distinct failure modes across T2I systems. Human studies confirm that our metrics correlate more strongly with human judgments than existing text-image metrics.

Current evaluation of German automatic text simplification (ATS) relies on general-purpose metrics such as SARI, BLEU, and BERTScore, which insufficiently capture simplification quality in terms of simplicity, meaning preservation, and fluency. While specialized metrics like LENS have been developed for English, corresponding efforts for German have lagged behind due to the absence of human-annotated corpora. To close this gap, we introduce DETECT, the first German-specific metric that holistically evaluates ATS quality across all three dimensions of simplicity, meaning preservation, and fluency, and is trained entirely on synthetic large language model (LLM) responses. Our approach adapts the LENS framework to German and extends it with (i) a pipeline for generating synthetic quality scores via LLMs, enabling dataset creation without human annotation, and (ii) an LLM-based refinement step for aligning grading criteria with simplification requirements. To the best of our knowledge, we also construct the largest German human evaluation dataset for text simplification to validate our metric directly. Experimental results show that DETECT achieves substantially higher correlations with human judgments than widely used ATS metrics, with particularly strong gains in meaning preservation and fluency. Beyond ATS, our findings highlight both the potential and the limitations of LLMs for automatic evaluation and provide transferable guidelines for general language accessibility tasks.

Many emerging applications of AI--from scientific discovery to medical diagnosis--require agents to seek information strategically: forming hypotheses, asking targeted questions, and making decisions under uncertainty. In high-stakes settings with limited resources, do language models (LMs) behave like rational agents? Drawing on insights from human cognition, we develop methods to evaluate and enhance agentic information-seeking. First, we introduce a decision-oriented dialogue task called Collaborative Battleship, in which a Captain must balance exploration (asking questions) and action (taking shots), while a Spotter must supply accurate, contextually-grounded answers. Compared to human players (N=42), we find that many LM agents struggle to ask informative questions, produce accurate answers, and identify high-utility actions. To address these gaps, we develop novel Monte Carlo inference strategies for LMs inspired by Bayesian Experimental Design (BED). For Spotter agents, our approach boosts accuracy by up to 14.7% absolute over LM-only baselines; for Captain agents, it raises expected information gain (EIG) by up to 0.227 bits (94.2% of the achievable noise ceiling). Combined, these components yield sharper targeting (+0.303-0.374 F1), and enable weaker LMs, such as Llama-4-Scout, to outperform both humans (8% -> 82% win rate) and frontier models (0% -> 67% win rate vs. GPT-5) at ~1% of GPT-5's cost. We replicate these findings on Guess Who?, where our methods significantly boost accuracy (+28.3-42.4 p.p.), demonstrating their general applicability for building information-seeking agents.

Masked autoregressive models (MAR) have emerged as a powerful paradigm for image and video generation, combining the flexibility of masked modeling with the expressiveness of continuous tokenizers. However, when sampling individual frames, video MAR models often produce highly distorted outputs due to the lack of a structured global prior, especially when using only a few sampling steps. To address this, we propose CanvasMAR, a novel autoregressive video prediction model that predicts high-fidelity frames with few sampling steps by introducing a canvas--a blurred, global one-step prediction of the next frame that serves as a non-uniform mask during masked generation. The canvas supplies global structure early in sampling, enabling faster and more coherent frame synthesis. To further stabilize autoregressive sampling, we propose an easy-to-hard curriculum via a motion-aware sampling order that synthesizes relatively stationary regions before attending to highly dynamic ones. We also integrate compositional classifier-free guidance that jointly strengthens the canvas and temporal conditioning to improve generation fidelity. Experiments on the BAIR, UCF-101, and Kinetics-600 benchmarks demonstrate that CanvasMAR produces higher-quality videos with fewer autoregressive steps. On the challenging Kinetics-600 dataset, CanvasMAR achieves remarkable performance among autoregressive models and rivals advanced diffusion-based methods.

Micro-benchmarking offers a solution to the often prohibitive time and cost of language model development: evaluate on a very small subset of existing benchmarks. Can these micro-benchmarks, however, rank models as consistently as the full benchmarks they replace? And can they rank models more consistently than selecting a random subset of data points? In many scenarios, we find that the answer is no. We introduce a meta-evaluation measure for micro-benchmarking which investigates how well a micro-benchmark can rank two models as a function of their performance difference on the full benchmark. This approach can determine which model pairs can be ranked correctly by a micro-benchmark, allowing for a finer-grained analysis of the trade-off between micro-benchmark size and reliability. Prior work has suggested selecting as few as 10 examples; we find that no micro-benchmarking method can consistently rank model pairs 3.5 points of accuracy apart on MMLU-Pro or 4 points apart on BIG-bench Hard. In order to consistently rank model pairs with relatively similar performances, we show that often as many as 250 examples must be selected, at which point random sampling is competitive with existing micro-benchmarking methods. When comparing only 8B instruction-tuned models on MMLU-Pro micro-benchmarks with 25 examples, we find that more than half of pairwise comparisons are not likely to be preserved. Our work provides actionable guidance for both micro-benchmark users and developers in navigating the trade-off between evaluation efficiency and reliability.

Pressure sensors are widely integrated into modern Heating, Ventilation and Air Conditioning (HVAC) systems. As they are sensitive to acoustic pressure, they can be a source of eavesdropping. We introduce HVAC-EAR, which reconstructs intelligible speech from low-resolution, noisy pressure data with two key contributions: (i) We achieve intelligible reconstruction from as low as 0.5 kHz sampling rate, surpassing prior work limited to hot word detection, by employing a complex-valued conformer with a Complex Unifed Attention Block to capture phoneme dependencies; (ii) We mitigate transient HVAC noise by reconstructing both magnitude and phase of missing frequencies. For the first time, evaluations on real-world HVAC deployments show significant intelligibility up to 1.2 m distance, raising novel privacy concerns.

Unmanned aerial vehicle (UAV) research requires the integration of cutting-edge technology into existing autopilot frameworks. This process can be arduous, requiring extensive resources, time, and detailed knowledge of the existing system. ROSplane is a lean, open-source fixed-wing autonomy stack built by researchers for researchers. It is designed to accelerate research by providing clearly defined interfaces with an easily modifiable framework. Built around ROS 2, ROSplane allows for rapid integration of low or high-level control, path planning, or estimation algorithms. A focus on lean, easily-understood code and extensive documentation lowers the barrier to entry for researchers. Recent developments to ROSplane improve its capacity to accelerate UAV research, including the transition from ROS 1 to ROS 2, enhanced estimation and control algorithms, increased modularity, and an improved aerodynamic modeling pipeline. This aerodynamic modeling pipeline significantly reduces the effort of transitioning from simulation to real-world testing without requiring costly system identification or computational fluid dynamics tools. ROSplane's architecture reduces the effort required to integrate new research tools and methods, expediting hardware experimentation.

ROSflight is a lean, open-source autopilot ecosystem for unmanned aerial vehicles (UAVs). Designed by researchers for researchers, it is built to lower the barrier to entry to UAV research and accelerate the transition from simulation to hardware experiments by maintaining a lean (not full-featured), well-documented, and modular codebase. This publication builds on previous treatments and describes significant additions to the architecture that improve the modularity and usability of ROSflight, including the transition from ROS 1 to ROS 2, supported hardware, low-level actuator mixing, and the simulation environment. We believe that these changes improve the usability of ROSflight and enable ROSflight to accelerate research in areas like advanced-air mobility. Hardware results are provided, showing that ROSflight is able to control a multirotor over a serial connection at 400 Hz while closing all control loops on the companion computer.

Human-like motion generation for robots often draws inspiration from biomechanical studies, which often categorize complex human motions into hierarchical taxonomies. While these taxonomies provide rich structural information about how movements relate to one another, this information is frequently overlooked in motion generation models, leading to a disconnect between the generated motions and their underlying hierarchical structure. This paper introduces the \ac{gphdm}, a novel approach that learns latent representations preserving both the hierarchical structure of motions and their temporal dynamics to ensure physical consistency. Our model achieves this by extending the dynamics prior of the Gaussian Process Dynamical Model (GPDM) to the hyperbolic manifold and integrating it with taxonomy-aware inductive biases. Building on this geometry- and taxonomy-aware frameworks, we propose three novel mechanisms for generating motions that are both taxonomically-structured and physically-consistent: two probabilistic recursive approaches and a method based on pullback-metric geodesics. Experiments on generating realistic motion sequences on the hand grasping taxonomy show that the proposed GPHDM faithfully encodes the underlying taxonomy and temporal dynamics, and it generates novel physically-consistent trajectories.

Simultaneous speech-to-text translation systems must balance translation quality with latency. Although quality evaluation is well established, latency measurement remains a challenge. Existing metrics produce inconsistent results, especially in short-form settings with artificial presegmentation. We present the first comprehensive meta-evaluation of latency metrics across language pairs and systems. We uncover a structural bias in current metrics related to segmentation. We introduce YAAL (Yet Another Average Lagging) for a more accurate short-form evaluation and LongYAAL for unsegmented audio. We propose SoftSegmenter, a resegmentation tool based on soft word-level alignment. We show that YAAL and LongYAAL, together with SoftSegmenter, outperform popular latency metrics, enabling more reliable assessments of short- and long-form simultaneous speech translation systems. We implement all artifacts within the OmniSTEval toolkit: https://github.com/pe-trik/OmniSTEval.

Sign language recognition (SLR) is a machine learning task aiming to identify signs in videos. Due to the scarcity of annotated data, unsupervised methods like contrastive learning have become promising in this field. They learn meaningful representations by pulling positive pairs (two augmented versions of the same instance) closer and pushing negative pairs (different from the positive pairs) apart. In SLR, in a sign video, only certain parts provide information that is truly useful for its recognition. Applying contrastive methods to SLR raises two issues: (i) contrastive learning methods treat all parts of a video in the same way, without taking into account the relevance of certain parts over others; (ii) shared movements between different signs make negative pairs highly similar, complicating sign discrimination. These issues lead to learning non-discriminative features for sign recognition and poor results in downstream tasks. In response, this paper proposes a self-supervised learning framework designed to learn meaningful representations for SLR. This framework consists of two key components designed to work together: (i) a new self-supervised approach with free-negative pairs; (ii) a new data augmentation technique. This approach shows a considerable gain in accuracy compared to several contrastive and self-supervised methods, across linear evaluation, semi-supervised learning, and transferability between sign languages.

Retrieval-Augmented Generation (RAG) has become a core paradigm in document question answering tasks. However, existing methods have limitations when dealing with multimodal documents: one category of methods relies on layout analysis and text extraction, which can only utilize explicit text information and struggle to capture images or unstructured content; the other category treats document segmentation as visual input and directly passes it to visual language models (VLMs) for processing, yet it ignores the semantic advantages of text, leading to suboptimal retrieval and generation results. To address these research gaps, we propose the Co-Modality-based RAG (CMRAG) framework, which can simultaneously leverage texts and images for more accurate retrieval and generation. Our framework includes two key components: (1) a Unified Encoding Model (UEM) that projects queries, parsed text, and images into a shared embedding space via triplet-based training, and (2) a Unified Co-Modality-informed Retrieval (UCMR) method that statistically normalizes similarity scores to effectively fuse cross-modal signals. To support research in this direction, we further construct and release a large-scale triplet dataset of (query, text, image) examples. Experiments demonstrate that our proposed framework consistently outperforms single-modality--based RAG in multiple visual document question-answering (VDQA) benchmarks. The findings of this paper show that integrating co-modality information into the RAG framework in a unified manner is an effective approach to improving the performance of complex VDQA systems.

We present a case study evaluating large language models (LLMs) with 128K-token context windows on a technical question answering (QA) task. Our benchmark is built on a user manual for an agricultural machine, available in English, French, and German. It simulates a cross-lingual information retrieval scenario where questions are posed in English against all three language versions of the manual. The evaluation focuses on realistic "needle-in-a-haystack" challenges and includes unanswerable questions to test for hallucinations. We compare nine long-context LLMs using direct prompting against three Retrieval-Augmented Generation (RAG) strategies (keyword, semantic, hybrid), with an LLM-as-a-judge for evaluation. Our findings for this specific manual show that Hybrid RAG consistently outperforms direct long-context prompting. Models like Gemini 2.5 Flash and the smaller Qwen 2.5 7B achieve high accuracy (over 85%) across all languages with RAG. This paper contributes a detailed analysis of LLM performance in a specialized industrial domain and an open framework for similar evaluations, highlighting practical trade-offs and challenges.

Vision Transformers naturally accommodate sparsity, yet standard tokenization methods confine features to discrete patch grids. This constraint prevents models from fully exploiting sparse regimes, forcing awkward compromises. We propose Subpixel Placement of Tokens (SPoT), a novel tokenization strategy that positions tokens continuously within images, effectively sidestepping grid-based limitations. With our proposed oracle-guided search, we uncover substantial performance gains achievable with ideal subpixel token positioning, drastically reducing the number of tokens necessary for accurate predictions during inference. SPoT provides a new direction for flexible, efficient, and interpretable ViT architectures, redefining sparsity as a strategic advantage rather than an imposed limitation.

The notable gap between user-provided and model-preferred prompts poses a significant challenge for generating high-quality images with text-to-image models, compelling the need for prompt engineering. Current studies on prompt engineering can effectively enhance the style and aesthetics of generated images. However, they often neglect the semantic alignment between generated images and user descriptions, resulting in visually appealing but content-wise unsatisfying outputs. In this work, we propose VisualPrompter, a novel training-free prompt engineering framework that refines user inputs to model-preferred sentences. VisualPrompter utilizes an automatic self-reflection module that identifies absent concepts in the generated images, followed by a target-specific prompt optimization mechanism that revises the prompts in a fine-grained manner. By deconstructing prompts, introducing new elements at the atomic semantic level, and then reassembling them, our framework is able to maintain semantic consistency and integrity throughout the optimization process. Extensive experiments demonstrate the effectiveness of VisualPrompter, which achieves new state-of-the-art performance on multiple benchmarks for text-image alignment evaluation. Additionally, our framework features a plug-and-play design, making it highly adaptable to various generative models. Our code is available at https://github.com/teheperinko541/VisualPrompter.

As large language models (LLMs) are deployed in safety-critical settings, it is essential to ensure that their responses comply with safety standards. Prior research has revealed that LLMs often fail to grasp the notion of safe behaviors, resulting in either unjustified refusals to harmless prompts or the generation of harmful content. While substantial efforts have been made to improve their robustness, existing defenses often rely on costly fine-tuning of model parameters or employ suboptimal heuristic techniques. In this work, we take a novel approach to safeguard LLMs by learning to adapt the system prompts in instruction-tuned LLMs. While LLMs are typically pre-trained to follow a fixed system prompt, we investigate the impact of tailoring the system prompt to each specific user input on the safety of the responses. To this end, we propose $\textbf{Sysformer}$, a trans$\textbf{former}$ model that updates an initial $\textbf{sys}$tem prompt to a more robust system prompt in the LLM input embedding space while attending to the user prompt. While keeping the LLM parameters frozen, the Sysformer is trained to refuse to respond to a set of harmful prompts while responding ideally to a set of safe ones. Through extensive experiments on $5$ LLMs from different families and $2$ recent benchmarks, we demonstrate that Sysformer can significantly enhance the robustness of LLMs, leading to upto $80\%$ gain in the refusal rate on harmful prompts while enhancing the compliance with the safe prompts by upto $90\%$. Results also generalize well to sophisticated jailbreaking attacks, making LLMs upto $100\%$ more robust against different attack strategies. We hope our findings lead to cheaper safeguarding of LLMs and motivate future investigations into designing variable system prompts.

Identifying inputs that trigger specific behaviours or latent features in language models could have a wide range of safety use cases. We investigate a class of methods capable of generating targeted, linguistically fluent inputs that activate specific latent features or elicit model behaviours. We formalise this approach as context modification and present ContextBench -- a benchmark with tasks assessing core method capabilities and potential safety applications. Our evaluation framework measures both elicitation strength (activation of latent features or behaviours) and linguistic fluency, highlighting how current state-of-the-art methods struggle to balance these objectives. We enhance Evolutionary Prompt Optimisation (EPO) with LLM-assistance and diffusion model inpainting, and demonstrate that these variants achieve state-of-the-art performance in balancing elicitation effectiveness and fluency.

Systems built on the Robot Operating System (ROS) are increasingly easy to assemble, yet hard to govern and reliably coordinate. Beyond the sheer number of subsystems involved, the difficulty stems from their diversity and interaction depth. In this paper, we use a compact heterogeneous robotic system (HeROS), combining mobile and manipulation capabilities, as a demonstration vehicle under dynamically changing tasks. Notably, all its subsystems are powered by ROS. The use of compatible interfaces and other ROS integration capabilities simplifies the construction of such systems. However, this only addresses part of the complexity: the semantic coherence and structural traceability are even more important for precise coordination and call for deliberate engineering methods. The Model-Based Systems Engineering (MBSE) discipline, which emerged from the experience of complexity management in large-scale engineering domains, offers the methodological foundations needed. Despite their strengths in complementary aspects of robotics systems engineering, the lack of a unified approach to integrate ROS and MBSE hinders the full potential of these tools. Motivated by the anticipated impact of such a synergy in robotics practice, we propose a structured methodology based on MeROS - a SysML metamodel created specifically to put the ROS-based systems into the focus of the MBSE workflow. As its methodological backbone, we adapt the well-known V-model to this context, illustrating how complex robotic systems can be designed with traceability and validation capabilities embedded into their lifecycle using practices familiar to engineering teams.

In clinical imaging, magnetic resonance (MR) image volumes are often acquired as stacks of 2D slices with decreased scan times, improved signal-to-noise ratio, and image contrasts unique to 2D MR pulse sequences. While this is sufficient for clinical evaluation, automated algorithms designed for 3D analysis perform poorly on multi-slice 2D MR volumes, especially those with thick slices and gaps between slices. Super-resolution (SR) methods aim to address this problem, but previous methods do not address all of the following: slice profile shape estimation, slice gap, domain shift, and non-integer or arbitrary upsampling factors. In this paper, we propose ECLARE (Efficient Cross-planar Learning for Anisotropic Resolution Enhancement), a self-SR method that addresses each of these factors. ECLARE uses a slice profile estimated from the multi-slice 2D MR volume, trains a network to learn the mapping from low-resolution to high-resolution in-plane patches from the same volume, and performs SR with anti-aliasing. We compared ECLARE to cubic B-spline interpolation, SMORE, and other contemporary SR methods. We used realistic and representative simulations so that quantitative performance against ground truth can be computed, and ECLARE outperformed all other methods in both signal recovery and downstream tasks. Importantly, as ECLARE does not use external training data it cannot suffer from domain shift between training and testing. Our code is open-source and available at https://www.github.com/sremedios/eclare.

We demonstrate the surprising real-world effectiveness of a very simple approach to whole-body model-predictive control (MPC) of quadruped and humanoid robots: the iterative LQR (iLQR) algorithm with MuJoCo dynamics and finite-difference approximated derivatives. Building upon the previous success of model-based behavior synthesis and control of locomotion and manipulation tasks with MuJoCo in simulation, we show that these policies can easily generalize to the real world with few sim-to-real considerations. Our baseline method achieves real-time whole-body MPC on a variety of hardware experiments, including dynamic quadruped locomotion, quadruped walking on two legs, and full-sized humanoid bipedal locomotion. We hope this easy-to-reproduce hardware baseline lowers the barrier to entry for real-world whole-body MPC research and contributes to accelerating research velocity in the community. Our code and experiment videos will be available online at:https://johnzhang3.github.io/mujoco_ilqr

In this paper, we introduce Context-Aware Priority Sampling (CAPS), a novel method designed to enhance data efficiency in learning-based autonomous driving systems. CAPS addresses the challenge of imbalanced datasets in imitation learning by leveraging Vector Quantized Variational Autoencoders (VQ-VAEs). In this way, we can get structured and interpretable data representations, which help to reveal meaningful patterns in the data. These patterns are used to group the data into clusters, with each sample being assigned a cluster ID. The cluster IDs are then used to re-balance the dataset, ensuring that rare yet valuable samples receive higher priority during training. We evaluate our method through closed-loop experiments in the CARLA simulator. The results on Bench2Drive scenarios demonstrate the effectiveness of CAPS in enhancing model generalization, with substantial improvements in both driving score and success rate.

Multi-modal large language model (MLLM) inference scheduling enables strong response quality under practical and heterogeneous budgets, beyond what a homogeneous single-backend setting can offer. Yet online MLLM task scheduling is nontrivial, as requests vary sharply in modality composition and latent reasoning difficulty, while execution backends incur distinct, time-varying costs due to system jitter and network variation. These coupled uncertainties pose two core challenges: deriving semantically faithful yet scheduling-relevant multi-modal task representations, and making low-overhead online decisions over irreversible multi-dimensional budgets. Accordingly, we propose \emph{M-CMAB} (\underline{M}ulti-modal \underline{M}ulti-constraint \underline{C}ontextual \underline{M}ulti-\underline{A}rmed \underline{B}andit), a multi-adapter-enhanced MLLM inference scheduling framework with three components: (i) a CLS-attentive, frozen-backbone \emph{Predictor} that extracts compact task representations and updates only lightweight adapters for action-specific estimation; (ii) a primal-dual \emph{Constrainer} that maintains online Lagrange multipliers to enforce long-horizon constraints via per-round objectives; and (iii) a two-phase \emph{Scheduler} that balances exploration and exploitation under irreversible budgets. We establish a regret guarantee under multi-dimensional knapsack constraints. On a composite multimodal benchmark with heterogeneous backends, \emph{M-CMAB} consistently outperforms state-of-the-art baselines across budget regimes, achieving up to 14.18% higher reward and closely tracking an oracle-aided upper bound. Codes are available at https://anonymous.4open.science/r/M2CMAB/.

Facial attribute classification relies on large-scale annotated datasets in which many traits, such as age and expression, are inherently ambiguous and continuous but are discretized into categorical labels. Annotation inconsistencies arise from subjectivity and visual confounders such as pose, illumination, expression, and demographic variation, creating mismatch between images and assigned labels. These inconsistencies introduce supervision errors that impair representation learning and degrade downstream prediction. We introduce DiffInf, a self-influence--guided diffusion framework for mitigating annotation inconsistencies in facial attribute learning. We first train a baseline classifier and compute sample-wise self-influence scores using a practical first-order approximation to identify training instances that disproportionately destabilize optimization. Instead of discarding these influential samples, we apply targeted generative correction via a latent diffusion autoencoder to better align visual content with assigned labels while preserving identity and realism. To enable differentiable guidance during correction, we train a lightweight predictor of high-influence membership and use it as a surrogate influence regularizer. The edited samples replace the originals, yielding an influence-refined dataset of unchanged size. Across multi-class facial attribute classification, DiffInf consistently improves generalization compared with standard noisy-label training, robust optimization baselines, and influence-based filtering. Our results demonstrate that repairing influential annotation inconsistencies at the image level enhances downstream facial attribute classification without sacrificing distributional coverage.

Large language models (LLMs) can now translate a researcher's plain-language goal into executable computation, yet scientific workflows demand determinism, provenance, and governance that are difficult to guarantee when an LLM decides what runs. Semi-structured interviews with 18 experts across 10 industrial R&D stakeholders surface 2 competing requirements--deterministic, constrained execution and conversational flexibility without workflow rigidity--together with boundary properties (human-in-the-loop control and transparency) that any resolution must satisfy. We propose schema-gated orchestration as the resolving principle: the schema becomes a mandatory execution boundary at the composed-workflow level, so that nothing runs unless the complete action--including cross-step dependencies--validates against a machine-checkable specification. We operationalize the 2 requirements as execution determinism (ED) and conversational flexibility (CF), and use these axes to review 20 systems spanning 5 architectural groups along a validation-scope spectrum. Scores are assigned via a multi-model protocol--15 independent sessions across 3 LLM families--yielding substantial-to-near-perfect inter-model agreement (Krippendorff a=0.80 for ED and a=0.98 for CF), demonstrating that multi-model LLM scoring can serve as a reusable alternative to human expert panels for architectural assessment. The resulting landscape reveals an empirical Pareto front--no reviewed system achieves both high flexibility and high determinism--but a convergence zone emerges between the generative and workflow-centric extremes. We argue that a schema-gated architecture, separating conversational from execution authority, is positioned to decouple this trade-off, and distill 3 operational principles--clarification-before-execution, constrained plan-act orchestration, and tool-to-workflow-level gating--to guide adoption.

Reassembling real-world archaeological artifacts from fragmented pieces poses significant challenges due to erosion, missing regions, irregular shapes, and large-scale ambiguity. Traditional jigsaw puzzle solvers, often designed for clean synthetic scenarios, struggle under these conditions, especially when the number of fragments grows into the thousands, as in the RePAIR benchmark. In this paper, we propose a human-in-the-loop (HIL) puzzle solving framework designed to address the complexity and scale of real-world cultural heritage reconstruction. Our approach integrates an automatic relaxation-labeling solver with interactive human guidance, allowing users to iteratively lock verified placements, correct errors, and guide the system toward semantically and geometrically coherent assemblies. We introduce two complementary interaction strategies, Iterative Anchoring and Continuous Interactive Refinement, which support scalable reconstruction across varying levels of ambiguity and puzzle size. Experiments on several RePAIR groups demonstrate that our hybrid approach substantially outperforms both fully automatic and manual baselines in accuracy and efficiency, offering a practical solution for large-scale expert-in-the-loop artifact reassembly.

Scene Graph Generation (SGG) is a task that encodes visual relationships between objects in images as graph structures. SGG shows significant promise as a foundational component for downstream tasks, such as reasoning for embodied agents. To enable real-time applications, SGG must address the trade-off between performance and inference speed. However, current methods tend to focus on one of the following: (1) improving relation prediction accuracy, (2) enhancing object detection accuracy, or (3) reducing latency, without aiming to balance all three objectives simultaneously. To address this limitation, we build on the powerful Real-time Efficiency and Accuracy Compromise for Tradeoffs in Scene Graph Generation (REACT) architecture and propose REACT++, a new state-of-the-art model for real-time SGG. By leveraging efficient feature extraction and subject-to-object cross-attention within the prototype space, REACT++ balances latency and representational power. REACT++ achieves the highest inference speed among existing SGG models, improving relation prediction accuracy without sacrificing object detection performance. Compared to the previous REACT version, REACT++ is 20% faster with a gain of 10% in relation prediction accuracy on average. The code is available at https://github.com/Maelic/SGG-Benchmark.

Foundation models such as Segment Anything Model 3 (SAM3) enable flexible text-guided medical image segmentation, yet their predictions remain highly sensitive to prompt formulation. Even semantically equivalent descriptions can yield inconsistent masks, limiting reliability in clinical and pathology workflows. We reformulate prompt sensitivity as a group-wise consistency problem. Semantically related prompts are organized into \emph{prompt groups} sharing the same ground-truth mask, and a prompt group-aware training framework is introduced for robust text-guided nuclei segmentation. The approach combines (i) a quality-guided group regularization that leverages segmentation loss as an implicit ranking signal, and (ii) a logit-level consistency constraint with a stop-gradient strategy to align predictions within each group. The method requires no architectural modification and leaves inference unchanged. Extensive experiments on multi-dataset nuclei benchmarks show consistent gains under textual prompting and markedly reduced performance variance across prompt quality levels. On six zero-shot cross-dataset tasks, our method improves Dice by an average of 2.16 points. These results demonstrate improved robustness and generalization for vision-language segmentation in computational pathology.

Accurate canopy height information is essential for quantifying forest carbon, monitoring restoration and degradation, and assessing habitat structure, yet high-fidelity measurements from airborne laser scanning (ALS) remain unevenly available globally. Here we present CHMv2, a global, meter-resolution canopy height map derived from high-resolution optical satellite imagery using a depth-estimation model built on DINOv3 and trained against ALS canopy height models. Compared to existing products, CHMv2 substantially improves accuracy, reduces bias in tall forests, and better preserves fine-scale structure such as canopy edges and gaps. These gains are enabled by a large expansion of geographically diverse training data, automated data curation and registration, and a loss formulation and data sampling strategy tailored to canopy height distributions. We validate CHMv2 against independent ALS test sets and against tens of millions of GEDI and ICESat-2 observations, demonstrating consistent performance across major forest biomes.

Whole-slide image (WSI) analysis is challenging due to the gigapixel scale of slides and their inherent hierarchical multi-resolution structure. Existing multiple instance learning (MIL) approaches often model WSIs as unordered collections of patches, which limits their ability to capture structured dependencies between global tissue organization and local cellular patterns. Although recent State Space Models (SSMs) enable efficient modeling of long sequences, how to structure WSI tokens to fully exploit their spatial hierarchy remains an open problem.We propose MoEMambaMIL, a structure-aware SSM framework for WSI analysis that integrates region-nested selective scanning with mixture-of-experts (MoE) modeling. Leveraging multi-resolution preprocessing, MoEMambaMIL organizes patch tokens into region-aware sequences that preserve spatial containment across resolutions. On top of this structured sequence, we decouple resolution-aware encoding and region-adaptive contextual modeling via a combination of static, resolution-specific experts and dynamic sparse experts with learned routing. This design enables efficient long-sequence modeling while promoting expert specialization across heterogeneous diagnostic patterns. Experiments demonstrate that MoEMambaMIL achieves the best performance across 9 downstream tasks.

We present Rewis3d, a framework that leverages recent advances in feed-forward 3D reconstruction to significantly improve weakly supervised semantic segmentation on 2D images. Obtaining dense, pixel-level annotations remains a costly bottleneck for training segmentation models. Alleviating this issue, sparse annotations offer an efficient weakly-supervised alternative. However, they still incur a performance gap. To address this, we introduce a novel approach that leverages 3D scene reconstruction as an auxiliary supervisory signal. Our key insight is that 3D geometric structure recovered from 2D videos provides strong cues that can propagate sparse annotations across entire scenes. Specifically, a dual student-teacher architecture enforces semantic consistency between 2D images and reconstructed 3D point clouds, using state-of-the-art feed-forward reconstruction to generate reliable geometric supervision. Extensive experiments demonstrate that Rewis3d achieves state-of-the-art performance in sparse supervision, outperforming existing approaches by 2-7% without requiring additional labels or inference overhead.