Federated Unlearning (FU) is an emerging paradigm in Federated Learning (FL) that enables participating clients to fully remove their contributions from a trained global model, driven by data protection regulations that mandate the right to be forgotten. However, existing FU methods mostly rely on synchronous coordination. This requirement forces the entire federation to halt and wait for stragglers to complete erasure, creating significant delays due to device heterogeneity. Furthermore, these methods often face the problem that the influence of erased data is merely suppressed temporarily and resurfaces during subsequent training, rather than being genuinely removed. To overcome these limitations, this paper proposes Asynchronous Federated Unlearning with Invariance Calibration (AFU-IC), a novel framework for medical imaging that decouples the erasure process from the global training workflow. This enables the target client to perform unlearning asynchronously without interrupting global training. Meanwhile, a server-side invariance calibration mechanism prevents the model from relearning the erased data. Extensive experiments on three medical benchmarks demonstrate that AFU-IC achieves unlearning efficacy and model fidelity comparable to gold-standard retraining while significantly reducing wall-clock latency compared to synchronous baselines. AFU-IC ensures efficient, compliant and reliable FL in cross-silo medical environments.

Despite being resource-intensive to train, 3D convolutional neural networks (CNNs) have been the standard approach to classify CT and MRI scans. Recent work suggests that deep multiple instance learning (MIL) may be a more efficient alternative for 3D brain scans, especially when the pre-trained image encoder used to embed each 2D slice is frozen and only the pooling operation and classifier are trained. In this paper, we provide a systematic comparison of simple MIL, attention-based MIL, 3D CNNs, and 3D ViTs across three CT and four MRI datasets, including two large datasets of at least 10,000 scans. Our goal is to help resource-constrained practitioners understand which neural networks work well for 3D neuroimages and why. We further compare design choices for attention-based MIL, including different encoders, pooling operations, and architectural orderings. We find that simple mean pooling MIL, without any learnable attention, matches or outperforms recent MIL or 3D CNN alternatives on 4 of 6 moderate-sized tasks. This baseline remains competitive on two large datasets while being 25x faster to train. To explain mean pooling's success, we examine per-slice attention quality and a semi-synthetic dataset where we can derive the best possible classifier via a Bayes estimator. This analysis reveals the limits of existing MIL approaches and suggests routes for future improvements.

Transformer-based architectures have established a dominant paradigm in global semantic perception; however, they remain fundamentally constrained by the profound spatial heterogeneity inherent in natural images. Specifically, the imposition of a uniform global receptive field across regions of varying information density inevitably leads to local feature degradation, particularly in dense conflict zones populated by microscopic targets. To address this mechanistic limitation, we propose ViCrop-Det, a training-free inference framework that introduces adaptive spatial trust region shrinkage. Inspired by the use of attention entropy in anomaly segmentation, ViCrop-Det leverages the detection decoder's cross-attention distribution as an endogenous probe. By utilizing Spatial Attention Entropy (SAE) to heuristically evaluate local spatial ambiguity, the framework executes dynamic spatial routing, allocating a fixed computational budget exclusively to regions exhibiting both high target saliency and high cognitive uncertainty. By shrinking the spatial trust region and injecting high-frequency localized observations, ViCrop-Det actively resolves spatial ambiguity and recovers fine-grained features without requiring architectural modifications. Extensive evaluations on VisDrone and DOTA-v1.5 demonstrate that ViCrop-Det yields competitive performance enhancements, consistently adding +1-3 mAP@50 to RT-DETR-R50 and Deformable DETR with a marginal 20-23\% latency overhead. On MS COCO, $AP_{S}$ improves while $AP_{M}/AP_{L}$ remains stable, indicating precise fine-scale refinement without compromising the global spatial prior. Under compute-matched settings, our adaptive routing strategy comprehensively surpasses uniform slicing baselines, achieving a highly optimized accuracy-speed trade-off.

We consider the problems of computing the optimal rank-$1$ Hankel and Toeplitz-structured approximation of arbitrary matrices under $L_2$ and $L_1$-norm error. Such problems arise naturally in engineered systems, including the basic few-shot signal Direction-of-Arrival (DoA) estimation problem that is of importance to modern autonomous systems applications. We develop accurate and computationally efficient structured matrix decomposition algorithms for both formulations and then derive analytically grounded small-sample-support DoA estimators for practical sensing system deployments. The resulting estimators under the $L_2$ and $L_1$ norms are formally shown to be maximum-likelihood optimal under white Gaussian and Laplace noise, respectively. The estimators are further validated through extensive simulation studies and real-world data experiments in few-shot DoA inference.

Surgical training involves didactic teaching, mentor-led learning, surgical skills laboratories, and direct exposure to surgery; however, increasing clinical pressures have limited operating room (OR) exposure. This work leverages virtual reality (VR) to provide a safe and immersive training environment. Existing VR training is often based on standardized scenarios not tailored to individual clinical cases. This study addresses this limitation using artificial intelligence (AI) based computer vision methods to generate patient-specific simulations from computed tomography (CT) and magnetic resonance imaging (MRI). This study focuses on patient-specific spinal decompression simulation for spinal stenosis in a virtual operating room. The objectives were (1) automatic creation of 3D anatomical models and (2) VR simulation of spinal decompression procedures including laminectomy, disc resection, and foraminotomy. Model construction required multimodal fusion (registration) of CT and MRI and segmentation of relevant structures. Segmentation was evaluated using the Dice Similarity Coefficient (DSC), and registration accuracy using Target Registration Error (TRE). Qualitative feedback was obtained from surgeons and trainees. High-fidelity patient-specific 3D models were generated efficiently (approximately 2.5 minutes per case, N = 15). Segmentation accuracy was high, with a DSC of 0.95 (+/- 0.03) for vertebral bone and 0.895 (+/- 0.02) for soft tissue structures. Registration accuracy showed a mean TRE of 1.73 (+/- 0.42) mm. Semi-structured interviews indicated improved spatial understanding, increased procedural confidence, and strong perceived educational value. This platform significantly reduced the time and costs of patient-specific modelling, thereby facilitating pre-operative planning, post-procedural assessments, and comprehensive surgical simulation.

RL post-training of frontier language models is increasingly bottlenecked by autoregressive rollout generation, making rollout acceleration a central systems challenge. Many existing efficiency methods improve throughput by changing the rollout or optimization regime, for example, through off-policy execution, replay, or lower-precision generation. We study speculative decoding as a lossless acceleration primitive for RL rollouts that preserves the target model's output distribution. We implement speculative decoding in NeMo-RL with a vLLM backend, supporting both synchronous and asynchronous pipelines and enabling speculation during RL rollouts. This benefit is realizable across speculation mechanisms, such as pretrained MTP heads, small external draft models or even techniques such as Eagle3, which are traditionally applied after RL phase. This yields a deployment path for state-of-the-art speculative decoding inside RL training. In a reasoning post-training workload at 8B scale under synchronous RL, speculative decoding improves rollout throughput by 1.8x. Using a high-fidelity performance simulator, we project that combining speculative decoding with asynchronous RL yields up to 2.5x end-to-end training speedup at 235B scale.

Open-vocabulary change detection aims to identify semantic changes in bi-temporal remote sensing images without predefined categories. Recent methods combine foundation models such as SAM, DINO and CLIP, but typically process each timestamp independently or interact only at the final comparison stage. Such paradigms suffer from insufficient temporal coupling during semantic reasoning, which limits their ability to distinguish genuine semantic changes from non-semantic appearance discrepancies. In addition, patch-dominant inference on high-resolution images often weakens global semantic continuity and produces fragmented change regions. To address these issues, we propose MemOVCD, a training-free open-vocabulary change detection framework based on cross-temporal memory reasoning and global-local adaptive rectification. Specifically, we reformulate bi-temporal change detection as a two-frame tracking problem and introduce weighted bidirectional propagation to aggregate semantic evidence from both temporal directions. To stabilize memory propagation across large temporal gaps, we construct histogram-aligned transition frames to smooth abrupt appearance changes. Moreover, a global-local adaptive rectification strategy adaptively fuses local and global-view predictions, improving spatial consistency while preserving fine-grained details. Experiments on five benchmarks demonstrate that MemOVCD achieves favorable performance on two change detection tasks, validating its effectiveness and generalization under diverse open-vocabulary settings.

Recent methods demonstrate that large-scale pretrained models, such as CLIP vision transformers, effectively detect AI-generated images (AIGIs) from unseen generative models when used as feature extractors. Many state-of-the-art methods for AI-generated image detection build upon the original CLIP-ViT to enhance this generalization. Since CLIP's release, numerous vision foundation models (VFMs) have emerged, incorporating architectural improvements and different training paradigms. Despite these advances, their potential for AIGI detection and AI image forensics remains largely unexplored. In this work, we present a comprehensive benchmark across multiple VFM families, covering diverse pretraining objectives, input resolutions, and model scales. We systematically evaluate their out-of-the-box performance for detecting fully-generated AI-images and AI-inpainted images, and discover that the best model outperforms the original CLIP by more than 12% in accuracy, beating established approaches in the process. To fully leverage the features of a modern VFM, we propose a simple redesign of the classifier head by utilizing tunable attention pooling (TAP), which aggregates output tokens into a refined global representation. Integrating TAP with the latest VFMs yields substantial performance gains across several AIGI detection benchmarks, establishing a new state-of-the-art on two challenging benchmarks for in-the-wild detection of AI-generated and -inpainted images.

Parametric Retrieval-Augmented Generation (PRAG) encodes external documents into lightweight parameter modules that can be retrieved and merged at inference time, offering a promising alternative to in-context retrieval augmentation. Despite its potential, many PRAG implementations train document adapters with task-supervised objectives, which may cause each adapter to encode both document-specific facts and reusable task-solving behavior. This entanglement may make adapter composition less reliable: when multiple adapters are merged at inference time, their overlapping task behaviors can accumulate together with document-specific updates, potentially making the merged adapter less stable and less focused on the intended document knowledge. To examine this issue, we explore Orthogonal Subspace Decomposition (OSD), an adapter-training setup that separates reusable task behavior from document-specific knowledge adapters. Concretely, we first train a Task LoRA to capture reusable task behavior, and then train document LoRAs to encode document-specific knowledge in a orthogonal subspace. This setup provides a controlled way to examine how orthogonalizing task and document LoRA updates affects adapter composition in multi-document PRAG. Experiments across multiple knowledge-intensive tasks and model scales suggest that this orthogonalization strategy can improve compositional robustness in parametric RAG, especially when multiple document adapters are merged.

Accurate and consistent Emergency Severity Index (ESI) assignment remains a persistent challenge in emergency departments, where highly variable free-text triage documentation contributes to mistriage and workflow inefficiencies. This study evaluates whether open-source small language models (SLMs) can serve as reliable, privacy-preserving decision-support tools for clinical triage. We systematically compared multiple SLMs across diverse prompting pipelines and found that clinical vignettes, concise summaries of triage narratives, yielded the most accurate predictions. The SLM, Qwen2.5-7B, demonstrated the strongest balance of accuracy, stability, and computational efficiency. Through large-scale domain adaptation using expert-curated and silver-standard pediatric triage data, fine-tuned Qwen2.5-7B models substantially reduced discordance and clinically significant errors, outperforming all baseline SLMs and advanced proprietary large language models (LLMs, e.g., GPT-4o). These findings highlight the feasibility of institution-specific SLMs for reliable, privacy-preserving ESI decision support and underscore the importance of targeted fine-tuning over more complex inference strategies.

The Probabilistic Transformer (PT) establishes that the Transformer's self-attention plus its feed-forward block is mathematically equivalent to Mean-Field Variational Inference (MFVI) on a Conditional Random Field (CRF). Under this equivalence the Transformer ceases to be a black-box neural network and becomes a programmable factor graph: graph topology, factor potentials, and the message-passing schedule are all explicit and inspectable primitives that can be engineered. PT was originally developed for natural language and in this report we investigate its potential for time series. We first lift PT into the Spatial-Temporal Probabilistic Transformer (ST-PT) to repair PT's missing channel axis and weak per-step semantics, and adopt ST-PT as a shared cornerstone backbone. We then identify three distinct properties that PT/ST-PT offers as a factor-graph model and derive three Research Questions, one per property, that probe how each property can be exploited in time series: RQ1. The graph topology and potentials are direct programmable primitives. Can this be used to inject symbolic time-series priors into ST-PT through structural graph modifications, especially under data scarcity and noise? RQ2. The CRF's factor matrices are the operator's potentials. Can an external condition program these factor matrices on a per-sample basis, so that conditional generation becomes structural rather than feature-level modulation of a fixed one? RQ3. Each MFVI iteration is a Bayesian posterior update on the factor graph. Can this turn the latent transition of latent-space AutoRegressive (AR) forecasting from an opaque MLP into a principled posterior update, and can a CRF teacher distill its latents into the AR student to counter cumulative error? We give one empirical study per question. Together, these three studies position ST-PT as a programmable framework for time-series modeling.

Languages of the world vary concerning the order of subject, object and verb. The most frequent dominant orders are SOV and SVO, and researchers have tailored models to this fact. However, there are still languages whose dominant order does not conform to these expectations or even lack a dominant order. Here we show that across linguistic families and macroareas, word order variation within languages is shaped by the principle of swap distance minimization even when the dominant order is not SOV/SVO and even when a dominant order is lacking.

Recent advances in self-evolution video understanding frameworks have demonstrated the potential of autonomous learning without human annotations. However, existing methods often suffer from weakly controlled optimization and uncontrolled difficulty progression, as they lack structured guidance throughout the iterative learning process. To address these limitations, we propose CurEvo, a curriculum-guided self-evolution framework that introduces curriculum learning into self-evolution to achieve more structured and progressive model improvement. CurEvo dynamically regulates task difficulty, refines evaluation criteria, and balances data diversity according to model competence, forming a curriculum-guided feedback loop that aligns learning complexity with model capability. Built upon this principle, we develop a multi-dimensional adaptive QA framework that jointly evolves question generation and answer evaluation across perception, recognition, and understanding dimensions, ensuring coherent and measurable curriculum progression. Through this integration, CurEvo transforms weakly controlled self-evolution into a more structured learning process for autonomous video understanding. Across seven backbones, CurEvo consistently improves both benchmark accuracy and evaluator-based semantic score on four VideoQA benchmarks, validating the effectiveness of curriculum-guided self-evolution for video understanding.

Skill libraries in deployed robotic systems are continually updated through fine-tuning, fresh demonstrations, or domain adaptation, yet existing typed-composition methods (BLADE, SymSkill, Generative Skill Chaining) treat the library as frozen at test time and do not analyze how composition outcomes change when a skill is replaced. We introduce a paired-sampling cross-version swap protocol on robosuite manipulation tasks to characterize this dimension of compositional skill learning. On a dual-arm peg-in-hole task we discover a dominant-skill effect: one ECM achieves 86.7% atomic success rate while every other ECM is at or below 26.7%, and whether this dominant ECM enters a composition shifts the success rate by up to +50pp. We characterize the boundary on a simpler pick task where all atomic policies saturate at 100% and the effect is undefined. Across three tasks we further find that off-policy behavioral distance metrics fail to identify the dominant ECM, ruling out the natural cheap predictor. We propose an atomic-quality probe and a Hybrid Selector combining per-skill probes (zero per-decision cost) with selective composition revalidation (full cost), and characterize its Pareto frontier on 144 skill-update decisions. On T6 the atomic-only probe sits 23pp below full revalidation (64.6% vs 87.5% oracle match) at zero per-decision cost; a Hybrid Selector with m=10 closes most of that gap to ~12pp at 46% of full-revalidation cost. On the cross-task average over 144 events, atomic-only is within 3pp of full revalidation under a mixed-oracle caveat. The atomic-quality probe is, to our knowledge, the first principled, deployment-ready primitive for skill-update governance in compositional robot policies.

Optical vibration sensing enables recovering the scene sound directly from the surface vibration of nearby objects, turning everyday objects into ``visual microphones''. However, most prior methods had focused on capturing the vibrations of specific objects with highly favorable vibration responses. These include objects where the surface vibrations are generated by the object itself (e.g., speaker membrane or guitar body) or objects consisting of a thin membrane which is highly reactive to sound (e.g., a chip bag or the leaf of a plant). In this paper, we tackle sound recovery for a more challenging class of solid objects whose vibration responses are poor or highly resonant. We simultaneously capture vibrations for multiple surface points on the object using a speckle-based vibrometry imaging system. Then, we derive a novel physics-guided vibration formation model that relates the scene sound source to the captured multi-point multi-axis vibrations via the object's vibrational modes. The model is then used to reverse the resonant transfer function of the vibrating object, fusing multiple vibration signals to estimate the original sound source in the scene. We evaluate our approach by recovering sound from a variety of everyday objects, demonstrating that it significantly outperforms traditional single-point speckle vibrometry in challenging scenarios and other signal-processing-based methods for multi-signal fusing.

We introduce a toolkit for uncovering spurious correlations between recording characteristics and target class in speech datasets. Spurious correlations may arise due to heterogeneous recording conditions, a common scenario for health-related datasets. When present both in the training and test data, these correlations result in an overestimation of the system performance -- a dangerous situation, specially in high-stakes application where systems are required to satisfy minimum performance requirements. Our toolkit implements a diagnostic method based on the detection of the target class using only the non-speech regions in the audio. Better than chance performance at this task indicates that information about the target class can be extracted from the non-speech regions, flagging the presence of spurious correlations. The toolkit is publicly available for research use.

Trust in clinical artificial intelligence (AI) cannot be reduced to model accuracy, fluency of generation, or overall positive user impression. In medicine, trust must be engineered as a measurable system property grounded in evidence, supervision, and operational boundaries of AI autonomy. This article proposes a practical framework for trustworthy clinical AI built around three principles: evidence, supervision, and staged autonomy. Rather than replacing deterministic clinical logic wholesale with end-to-end black-box models, the proposed approach combines a deterministic core, a patient-specific AI assistant for contextual validation, a multi-tier model escalation mechanism, and a human supervision layer for verification, escalation, and risk control. We demonstrate that trust also depends on selective verification of clinically critical findings, bounded clinical context, disciplined prompt architecture, and careful evaluation on realistic cases. Classifier-driven modular prompting is examined as an incremental path to scaling clinical depth without sacrificing prompt performance and without waiting for complete rule-based coverage. To operationalize trust, a set of trust metrics is proposed, built on metrological principles -- measurement uncertainty, calibration, traceability -- enabling quantitative rather than subjective assessment of each architectural layer. In this perspective, trustworthy clinical AI emerges not as a property of an individual model, but as an architectural outcome of a system into which evidence trails, human oversight, tiered escalation, and graduated action rights are embedded from the outset.

Conventional wavelet-domain methods for room impulse response denoising rely on thresholding detail coefficients, which is unsuited for low frequencies. In this work, we introduce a wavelet-based post-processing algorithm that extends denoising to approximation coefficients by means of sparse dictionary learning with a time-varying error tolerance. The proposed method leverages an exponential decay envelope model to adapt reconstruction accuracy according to the local signal-to-noise ratio. This approach significantly improves low-frequency denoising of synthetic and measured room impulse responses compared to the baseline method, leading to more accurate estimation of acoustic parameters such as decay time.

Differential Privacy (DP) for text matured from disjointed word-level substitutions to contiguous sentence-level rewriting by leveraging the generative capacity of language models. While this form of text privatization is best suited for balancing formal privacy guarantees with grammatical coherence, its impact on the register identity of text remains largely unexplored. By conducting a multidimensional stylistic profiling of differentially-private rewriting, we demonstrate that the cost of privacy extends far beyond lexical variation. Specifically, we find that rewriting under privacy constraints induces a systematic functional mutation of the text's communicative signature. This shift is characterized by the severe attrition of interactive markers, contextual references, and complex subordination. By comparing autoregressive paraphrasing against bidirectional substitution across a spectrum of privacy budgets, we observe that both architectures force convergence toward a non-involved and non-persuasive register. This register-blind sanitization effectively preserves semantic content but structurally homogenizes the nuanced stylistic markers that define human-authored discourse.

Large Reasoning Models (LRMs) achieve strong performance on mathematical reasoning tasks but remain unreliable on challenging instances. Existing test-time scaling methods, such as repeated sampling, self-correction, and tree search, improve performance at the cost of increased computation, yet often exhibit diminishing returns on hard problems. We observe that output disagreement is strongly correlated with instance difficulty and prediction correctness, providing a useful signal for guiding instance-level strategy selection at test time. Based on this insight, we propose a training-free framework that formulates test-time scaling as an instance-level routing problem, rather than allocating more computation within a single strategy, dynamically selecting among different scaling strategies based on output disagreement. The framework applies lightweight resolution for consistent cases, majority voting for moderate disagreement, and rewriting-based reformulation for highly ambiguous instances. Experiments on seven mathematical benchmarks and three models show that our method improves accuracy by 3% - 7% while reducing sampling cost compared to existing approaches.

Annotating long-horizon robotic demonstrations with precise temporal action boundaries is crucial for training and evaluating action segmentation and manipulation policy learning methods. Existing annotation tools, however, are often limited: they are designed primarily for vision-only data, do not natively support synchronized visualization of robot-specific time-series signals (e.g., gripper state or force/torque), or require substantial effort to adapt to different dataset formats. In this paper, we introduce ATLAS, an annotation tool tailored for long-horizon robotic action segmentation. ATLAS provides time-synchronized visualization of multi-modal robotic data, including multi-view video and proprioceptive signals, and supports annotation of action boundaries, action labels, and task outcomes. The tool natively handles widely used robotics dataset formats such as ROS bags and the Reinforcement Learning Dataset (RLDS) format, and provides direct support for specific datasets such as REASSEMBLE. ATLAS can be easily extended to new formats via a modular dataset abstraction layer. Its keyboard-centric interface minimizes annotation effort and improves efficiency. In experiments on a contact-rich assembly task, ATLAS reduced the average per-action annotation time by at least 6% compared to ELAN, while the inclusion of time-series data improved temporal alignment with expert annotations by more than 2.8% and decreased boundary error fivefold compared to vision-only annotation tools.

Deploying production-ready multi-agent systems (MAS) in complex industrial environments remains challenging due to limitations in scalability, observability, and autonomous evolution. We present OxyGent, an open-source framework driven by two core novelties: a unified Oxy abstraction and the OxyBank evolution engine. The unified abstraction encapsulates agents, tools, LLMs, and reasoning flows as pluggable atomic components, enabling Lego-like scalable system composition and non-intrusive monitoring. To enhance observability, OxyGent introduces permission-driven dynamic planning that replaces rigid workflows with execution graphs generated at runtime, providing adaptive visualizations. Furthermore, to support continuous evolution, OxyBank serves as an AI asset management platform that drives automated data backflow, annotation, and joint evolution. Empirical evaluations and real-world case studies show that OxyGent provides a robust and scalable foundation for MAS. OxyGent is fully open-sourced under the Apache License 2.0 at https://github.com/jd-opensource/OxyGent.

Retrieval-Augmented Generation (RAG) models frequently produce answers grounded in parametric memory rather than the retrieved context, undermining the core promise of retrieval augmentation. A fundamental obstacle to fixing this unfaithfulness is the lack of training data that explicitly requires models to prefer context over internal knowledge. We introduce Faithfulness-QA, a large-scale dataset of 99,094 samples constructed through counterfactual entity substitution. Starting from two established extractive QA benchmarks--SQuAD and TriviaQA--we automatically identify answer-bearing named entities in each context, replace them with type-consistent alternatives drawn from a curated bank of 76,953 entities, and thereby manufacture controlled knowledge conflicts between context and parametric memory. Rigorous quality filtering ensures 100% pass rates across four automated checks on random 200-sample audits. We release the full dataset, the construction pipeline, and a typed entity bank covering eight named entity categories. Faithfulness-QA is designed as a training resource for attention-based faithfulness objectives and as an evaluation benchmark for measuring context-grounding behavior in RAG systems. Data and code are available at https://github.com/qzhangFDU/faithfulness-qa-dataset.

Dimensionality reduction methods such as UMAP and t-SNE are central tools for visualising high-dimensional data, but their local-neighborhood objectives can preserve sampling noise while distorting global topology. We show that standard local metrics reward this noise memorisation: top-performing embeddings invent cycles and disconnected islands absent from the data. We introduce a topology-faithfulness benchmark based on noisy manifolds with known homology, tune DiRe against it, and find Pareto-optimal configurations that match or beat GPU-accelerated UMAP on classification while recovering exact first Betti numbers on stress tests. On 723K arXiv paper embeddings, DiRe preserves 3-4 times more topological structure than UMAP at comparable wall-clock.

A multilingual collection may contain useful knowledge in other languages to supplement and correct the facts in the original language for Retrieval-Augmented Generation (RAG). However, the vanilla approach that simply concatenates multiple pieces of knowledge from different languages into the context may fail to improve effectiveness due to the potential disparities across languages. To better leverage multilingual knowledge, we propose CroSearch-R1, a search-augmented reinforcement learning framework to integrate multilingual knowledge into the Group Relative Policy Optimization (GRPO) process. In particular, the approach adopts a multi-turn retrieval strategy with cross-lingual knowledge integration to dynamically align the knowledge from other languages as supplementary evidence into a unified representation space. Furthermore, we introduce a multilingual rollout mechanism to optimize reasoning transferability across languages. Experimental results demonstrate that our framework effectively leverages cross-lingual complementarity and improves the effectiveness of RAG with multilingual collections.

Research on video-based eye-tracking has long explored stereo and glint-based methods, yet existing wearable eye trackers - both commercial and open-source - offer limited flexibility for algorithm development and comparative evaluation. We present an affordable, wearable stereo eye-tracking platform built from off-the-shelf and 3D-printable components that explicitly targets this gap. The system combines four infrared eye cameras, infrared illumination, an optional scene camera, and software support for calibration and synchronized data acquisition. By design, the platform supports multiple eye-tracking paradigms, including stereo, glint-based, and binocular approaches, within a single hardware configuration. Rather than optimizing for end-user robustness, the platform prioritizes modularity and extensibility for research use. This paper focuses on the hardware architecture and calibration pipeline and demonstrates the feasibility of the approach using a prototype implementation. All hardware designs and documentation are made openly available.

Foundation models of brain activity promise a new frontier for in silico neuroscience by emulating neural responses to complex stimuli across tasks and modalities. A natural next step is to ask whether these models can also be used in reverse. Can we recover a stimulus or its properties from synthetic brain activity? We study this question in a proof-of-concept setting using TRIBEv2. We pair the brain emulator with large language models (LLMs) that generate news headlines from linguistic parameters such as valence, arousal, and dominance. We then use simulation-based inference to learn a probabilistic mapping from brain maps to latent stimulus parameters. Our results show that these parameters can be recovered from predicted brain maps, validating the quality of neural encodings. They also show that LLMs can serve as controllable stimulus generators for simulated experiments. Together, these findings provide a step toward decoding and inverse design with foundation brain models.

The discovery of novel materials is critical for global energy and quantum technology transitions. While deep learning has fundamentally reshaped this landscape, existing predictive or generative models typically operate in isolation, lacking the autonomous orchestration required to execute the full discovery process. Here we present ElementsClaw, an agentic framework for materials discovery that synergizes Large Atomic Models (LAMs) with Large Language Models (LLMs). In response to varied human queries, ElementsClaw orchestrates a suite of LAM tools finetuned from our proposed 1-billion-parameter model Elements for atomic-scale numerical computation, while leveraging LLMs for high-level semantic reasoning. This shift moves AI-driven materials science from isolated processes toward integrated and human interactive discovery. Applied to superconductors, ElementsClaw screens 2.4 million crystals in just 28 GPU hours to identify 68,000 high-confidence candidates (The complete dataset of screened superconductors is available at https://developer.damo-academy.com/material), expanding known superconducting space by orders of magnitude compared to datasets curated over decades. Critically, ElementsClaw achieves a high success rate in identifying superconductors hidden in literature and discovers four novel experimentally verified superconductors, exemplified by Zr3ScRe8 with a transition temperature of 6.8 K and HfZrRe4 at 6.7 K. Together, our results establish a knowledge integrated, autonomously orchestrated, and experimentally grounded paradigm for materials discovery.

The wide adoption of AI agents in complex human workflows is driving rapid growth in LLM token consumption. When agents are deployed on tasks that require a significant amount of tokens, three questions naturally arise: (1) Where do AI agents spend the tokens? (2) Which models are more token-efficient? and (3) Can agents predict their token usage before task execution? In this paper, we present the first systematic study of token consumption patterns in agentic coding tasks. We analyze trajectories from eight frontier LLMs on SWE-bench Verified and evaluate models' ability to predict their own token costs before task execution. We find that: (1) agentic tasks are uniquely expensive, consuming 1000x more tokens than code reasoning and code chat, with input tokens rather than output tokens driving the overall cost; (2) token usage is highly variable and inherently stochastic: runs on the same task can differ by up to 30x in total tokens, and higher token usage does not translate into higher accuracy; instead, accuracy often peaks at intermediate cost and saturates at higher costs; (3) models vary substantially in token efficiency: on the same tasks, Kimi-K2 and Claude-Sonnet-4.5, on average, consume over 1.5 million more tokens than GPT-5; (4) task difficulty rated by human experts only weakly aligns with actual token costs, revealing a fundamental gap between human-perceived complexity and the computational effort agents actually expend; and (5) frontier models fail to accurately predict their own token usage (with weak-to-moderate correlations, up to 0.39) and systematically underestimate real token costs. Our study offers new insights into the economics of AI agents and can inspire future research in this direction.

Neuron labeling assigns textual descriptions to internal units of deep networks. Existing approaches typically rely on highly activating examples, often yielding broad or misleading labels by focusing on dominant but incidental visual factors. Prior work such as FALCON introduced contrastive examples -- inputs that are semantically similar to activating examples but elicit low activations -- to sharpen explanations, but it primarily addresses subspace-level interpretability rather than scalable neuron-level labeling. We revisit contrastive explanations for neuron-level labeling in two stages: (1) candidate label generation with vision language models (VLMs) and (2) label assignment with CLIP-like encoders. First, we show that providing contrastive image sets to VLMs yields candidate labels that are more specific and more faithful. Second, we introduce Contrastive Semantic Projection (CSP), an extension of SemanticLens that incorporates contrastive examples directly into its CLIP-based scoring and selection pipeline. Across extensive experiments and a case study on melanoma detection, contrastive labeling improves both faithfulness and semantic granularity over state-of-the-art baselines. Our results demonstrate that contrastive examples are a simple yet powerful and currently underutilized component of neuron labeling and analysis pipelines.