Search research

Video world models are increasingly used as simulators, yet visual fidelity alone does not show that a model maintains the hidden state of the world. We examine this gap with an action-conditioned video Shell Game, a visual analog of $S_5$ state tracking that decouples visual rendering from compositing the hidden state underneath. Bidirectional and autoregressive Transformers, Mamba, and linear attention restricted to nonnegative transition eigenvalues all fit the training horizon of 5 swaps and then fall toward chance on longer swap chains (extrapolation) while still rendering plausible video with additional denoising steps providing no benefit. The pixel-based diffusion target never supervises the unseen hidden state, so the generated frames cannot carry it and the state has to live inside the architecture rather than in the tokens. For a Transformer, that architectural state is only an append-only KV cache, so the model has to re-derive the hidden arrangement from the whole history at every chunk. We find two mechanisms that do extrapolate, and both carry a state across chunks and revise it in place. Linear attention succeeds once its transition eigenvalues may be negative, and TTT with a nonlinear fast weight succeeds by updating the feature map through which it reads its own state. We further examine harder cases in dynamic world exploration tasks, and discuss the broader implications for building stateful video world models.

Traffic video understanding has become an important problem in intelligent transportation, as road videos provide direct evidence for accidents, violations, and interactions between vehicles and vulnerable road users. A useful system should explain how a traffic event develops, why it happens, and when the relevant interaction occurs, yet this remains difficult for multimodal large language models (MLLMs) because traffic videos contain sparse events and varied viewpoints. We introduce UniTraffic-Agent, the MR-CAS solution for Track~3 of the 10th AI City Challenge, which includes Traffic Anomaly Reasoning (TAR) and two out-of-domain evaluations: FETV for fisheye traffic events and PSI-VQA for pedestrian intention reasoning. UniTraffic-Agent follows an observe--reason--act--verify workflow that samples timestamped visual evidence, reasons over all questions from the same clip in one request, and converts responses through task-specific action adapters. On the official Public leaderboards, MR-CAS ranks 16th on TAR with a score of 0.5780, 2nd on FETV with 0.4884, and 4th on PSI-VQA with 64.4161. The code is available at https://github.com/Roclp/UniTraffic-Agent.

We present the first-place solution to the MeViS-Text track of the 8th Large-scale Video Object Segmentation (LSVOS) Challenge 2026: referring video object segmentation guided by written motion expressions, including deceptive no-target expressions that match no object in the video and must yield empty masks in every frame. Our pipeline, SSUPER, resolves each expression into a visual concept, generates full-video candidate masklets with SAM~3.1, and selects target IDs. At every reasoning stage, three heterogeneous multimodal large language models independently execute the same stage-specific prompt before a single synthesis pass commits one schema-validated verdict. Although this system rejects every no-target expression in validation, the leaderboard reveals that a substantial share of test no-target cases still slips through. The reason is that hard negatives name a plausible object and fail only under the complete temporal predicate, so when selection and existence are decided together, a category-plausible masklet anchors the verdict. Hence, we decouple existence verification into an independent multi-agent audit of the full predicate (category, count, action, trajectory, event order, and semantic role) that distinguishes absence from temporary invisibility, discounts apparent motion caused by camera movement, and requires contradicting evidence rather than mere uncertainty for a no-target verdict. Without any new segmentation call, this audit recovers most of the residual no-target errors. A training-data-only StyleRefiner then aligns mask geometry with the annotation style of MeViSv2 while preserving every presence decision by construction, showing that once the semantics are fixed, part of the remaining error is stylistic rather than semantic. The complete system reaches a Final score of 0.9081339614 on the official challenge leaderboard.

Understanding dynamic sound sources requires jointly determining what produces a sound, where the source is located, and how it moves over time. Yet existing audio-language models often represent clips as global acoustic events, while vision-language models lack the spatial audio cues needed to localize and track individual sources. To evaluate this missing capability, we introduce ST-OmniQA, a spatio-temporal audio-visual question-answering benchmark built from panoramic videos paired with synchronized first-order Ambisonics (FOA) audio of moving sound sources. It contains 40K videos and 400K question-answer pairs organized into four capability levels covering sound-event recognition, direction of arrival, source distance, motion trajectories, and temporally grounded audio-visual reasoning. Building on this benchmark, we propose ST-Omni-R1, which integrates FOA-derived semantic and trajectory representations with panoramic visual context and is trained through progressive curriculum learning and reasoning-tree reinforcement learning. ST-Omni-R1 achieves 77.83% average semantic accuracy across the four levels, compared with 37.28% for the best evaluated baseline. Results on three public spatial-audio benchmarks further indicate that its learned spatial and motion representations transfer beyond ST-OmniQA.

In Formula 1, drivers optimize racing lines within tire grip limits to minimize lap times; however, in rally racing, drivers intentionally break traction to drift on loose surfaces. This maneuver rapidly aligns the vehicle for corner exits, ultimately reducing lap time. Autonomously executing such maneuvers formulates a complex dual-objective control problem: stabilizing highly nonlinear drift dynamics while strictly minimizing lap time. Addressing this challenge motivates the development of advanced Minimum-Lap-Time (MLT) drift control architectures. This paper proposes a planning-control framework specifically designed for MLT drifting scenario. First, we formulate an optimal control problem to generate a MLT drift planning trajectory, which is used as prior data to train a deep reinforcement learning drift controller. Given that drifting involves extremely large sideslip angles and is therefore challenging to learn directly, a Track-guided Reinforcement Learning (TgRL) drift control method is proposed to enable progressive training in a step-by-step manner, from drift control policy, to drift corner policy, and finally to a comprehensive drift race policy. The reward function incorporates both an instant reward term and an end reward term derived from the Minimum-Lap-Time objective. Simulation results demonstrate that the proposed framework enables the agent to learn a drift racing policy that not only ensures vehicle motion control performance but also effectively reduces lap time.

This technical report presents our approach to Challenge Track~3: SeePhys Pro at the 3rd AI for Math Workshop, where the task is to answer college-level physics questions whose statement and figure may be given partly or entirely as an image. Visual physics problems become substantially harder for large language models when the decisive information resides in a figure rather than in the text, and this modality gap widens as more of the problem migrates into the image. We address the task with a two-stage framework: a visual information extraction stage that re-expresses figure content as solver-readable text to close the modality gap, and a reasoning stage that orchestrates three heterogeneous solvers through multi-agent debate. Our analysis yields two findings: the gain from orchestration comes from reliable answer selection rather than from additional debate, and the value of a figure aid scales with how much of the problem is locked inside the image. The resulting pipeline improves overall accuracy over a single-agent baseline from 0.643 to 0.802 on the public split, and won 1st place on both the public and the private leaderboard (private overall 0.743).

Memory-efficient optimizers such as GaLore train large language models by projecting gradients onto a rank-r subspace recomputed every T steps, assuming this subspace is a slowly drifting object that can be tracked. We show that beyond a small reproducible core, there is no such object. Two estimates of the top-r subspace computed at the same step from disjoint minibatches disagree as much as estimates computed T steps apart (0.73 vs 0.74 of the maximal chordal distance sqrt(2r), at Pythia-160M with r=128): the apparent rotation at each refresh is dominated by estimator noise. This holds across four model families in three architecture classes from 70M to 6.9B parameters, strengthening with scale, and more weakly in a vision transformer. Only ~39 of 128 directions are reproducible across minibatches, and averaging cannot recover the rest: under N-fold averaging the gradient's spectral tail shrinks as N^(-1/4) rather than the N^(-1/2) of pure noise, so no averaging budget makes the subspace well defined. What helps instead follows from treating each refresh as a change of coordinates for Adam's state. Carrying the second moment blindly is provably about (r-k*)/2 worse than the best rotation-blind estimator, while the first moment transports exactly through the rotation, the optimal linear map under isotropic gradients and the rule LDAdam uses. At 1B over 40k steps (3 seeds), full LDAdam reaches 18.7 perplexity at beta2=0.999, beating untransported GaLore after its best beta2 fix (19.3); shortening the second-moment memory to beta2=0.99 helps the refreshing optimizers, though for canonical GaLore the effect is small and a full-rank control reverses it. One measurable fact, subspace non-identifiability, clarifies why GaLore works, which patches work, and what to check before trusting a low-rank assumption: the reproducible rank k*.

Predicting thermal volatility in high-performance EV powertrains is difficult as internal temperatures are rarely observable outside the lab, and models calibrated on lab drive cycles fail when deployed against real-world loads. We study this lab-to-track transfer problem using conformal prediction, offering distribution-free uncertainty bounds. We implement Ensemble Batch Prediction Intervals (EnbPI; Xu & Xie, 2021), a leave-one-out bootstrap-ensemble conformal method for autocorrelated time series, and calibrate it on real CALCE lithium-ion cycler data (A123 SP20 cells, FUDS profile). We evaluate it under a genuine, measured covariate shift: a second real CALCE test condition (US06 Highway Driving Schedule at 45°C). The unweighted EnbPI bound, achieving its nominal 95% coverage in-distribution (measured: 95.00%), degrades to 70.13% empirical coverage under this real shift. We introduce a weighted EnbPI procedure combining EnbPI's ensemble residuals with density-ratio weighting (Tibshirani et al., 2019), estimating the density ratio via a probabilistic domain classifier. This recovers coverage to 72.42%, a modest, honestly-reported improvement, not a complete fix. We additionally apply the calibrated model to real 2023 Formula 1 telemetry (Monza and Silverstone, driver VER) as an unsupervised out-of-distribution diagnostic. Because no internal thermal channel exists in public trackside telemetry, we report only unsupervised flag rates (65.6% at Monza, 58.0% at Silverstone, well above the 5% in-distribution base rate) and note inconsistent associations between flags and braking/DRS zones. We conclude that conformal domain adaptation is a promising but only partially solved tool for this problem, detailing exactly where it falls short.

Visual Object Tracking (VOT) and Moving Object Segmentation (MOS) are two fundamental tasks in computer vision that involve both spatial and temporal object dynamics. Existing methods rely predominantly on visual cues and thus often falter in real-world scenarios where object motions are inherently complex and nonlinear. To address this limitation, we propose SUMO, a zero-shot, training-free, unified framework integrating nonlinear dynamics with vision-based segmentation for accurate and consistent VOT and MOS. Specifically, we develop a nonlinear State Space Model (SSM) inspired by robotics principles to capture the complex object dynamics. Building on this model, we propose a Selective Unscented Filter (SUF) for accurate state estimation, which features a joint scoring mechanism and dynamically fuses multi-source predictions to identify the most plausible object state over time. Furthermore, we apply a memory selection mechanism to evaluate the reliability of memory frames. Our extensive experimental results show that SUMO achieves state-of-the-art performance on both VOT and MOS tasks.

Confidence is an estimate of the probability that a chosen answer is correct. Verbal confidence reports are widely used as uncertainty measures in large language models, but whether they are best understood as estimates of correctness is unclear. We test this with a two-stage abstention paradigm from the neuroscience of perceptual decision making: a model first answers and reports its confidence, then decides whether to commit it to a user or abstain. Across four non-reasoning models, prompt framings, and confidence formats, verbal confidence predicted the commit/abstain decision substantially better than whether the answer was correct. Calibrated token log-probabilities showed the opposite profile, with abstention-prediction coupled to correctness discrimination, the signature of an answer-evidence signal. After removing the variance verbal confidence shared with log-probabilities, the residual stayed aligned with commitment while its link to correctness fell to near chance. The dissociation generalised to four reasoning models across four benchmarks of varying difficulty, from hard multiple-choice to frontier-level freeform questions. Mechanistic analyses in Gemma 3 and 4 were convergent: a post-answer state known to causally support verbal-confidence generation already encoded the future abstention decision before the abstention prompt, organised mainly by that decision rather than by correctness, the two lying in approximately orthogonal directions in activation space. Steering along a verbal-confidence-specific direction causally shifted abstention. Verbal and log-probability confidence are thus not interchangeable: log-probabilities track answer evidence and correctness, whereas verbal confidence is better understood as a behaviour-facing readout of an internal commit-readiness state, challenging the practice of treating verbal reports as proxies for reliability.

4D generation aims to animate 3D objects with realistic motion, holding great promise for applications. Existing methods typically decouple 3D asset generation from motion synthesis: acquire a 3D asset, prepare a structural representation like mesh and Gaussians, and synthesize motion from text or video control signals. However, dense mesh and Gaussian representations incur high computational costs and are prone to temporal artifacts, limiting animation quality and duration to only short clips. Meanwhile, text lacks fine-grained spatial and temporal details such as timing and coordination, while video entangles motion with appearance and background. Together, these limitations result in 4D animations that suffer from poor temporal consistency, wrong identification, and limited controllability. We address these issues with \texttt{ACT}, a trajectory-conditioned framework for topology-general skeletal animation. ACT uses skeletons as a compact structured and compute-efficient representation and 3D point trajectories from monocular video as explicit motion guidance which provide detailed motion patterns without appearance entanglement. At the core of ACT is a Routed Trajectory Injector, which achieves accurate and robust trajectory-to-joint transfer through three complementary designs: prior-guided hard routing establishes precise skeleton-to-mesh correspondences, global routing enables holistic joint-track interaction for full-body motion awareness, and local windowed cross-attention enforces fine-grained temporal alignment, improving micro-timing and reducing motion misalignment across varying motion rates. Extensive experiments demonstrate that \texttt{ACT} significantly outperforms existing methods in fidelity and temporal consistency.

We investigate whether language models internally track the value of their current trajectory, defined as the likelihood that their ongoing strategy will achieve their goals. Using synthetic, in-context reinforcement learning data, we construct a "value" axis for Qwen3-8B. We find that activations along this axis distinguish between high vs. low verbalized confidence, rollouts without and with backtracking, and correct vs. corrupted code. Steering towards high value causally suppresses self-correction and reduces explanatory verbosity, while steering towards low value induces backtracking and exploration. We demonstrate that direct preference optimization (DPO) can increase the internal value of rewarded behaviors (e.g. use a certain word), causing the model to act more confidently after exhibiting them. Finally, we apply the value axis to study in-the-wild settings. For example, we find that Qwen assigns low value to politically sensitive chat queries after post-training and that supervised fine-tuning increases internal confidence within the training domain. Our results suggest that language models linearly encode an estimate of expected goal success that modulates their confidence in pursuing a direction.

Accurate and computationally efficient vehicle models are essential for motion planning of autonomous vehicles, where positional accuracy directly affects trajectory feasibility and safety. However, the positional accuracy has not been systematically evaluated against real measurements. Therefore, this paper compares the short-horizon positional accuracy of three single-track vehicle models against vehicle measurements across various driving maneuvers. Model parameters are identified through dedicated experiments with the instrumented test vehicle. Rather than identifying a single best model, this work aims to provide insight into the trade-offs between model complexity, parameterization quality, and positional accuracy for informed model selection in Model Predictive Control applications.

Evaluation in machine learning is typically treated as a neutral measurement process. However, in operational information systems, evaluation outcomes are often conditioned by the processes used to generate labels. This paper does not seek to improve classification performance. Instead, it examines the validity of performance measurement under differing label-authority regimes. This issue is particularly relevant in large-scale metadata-driven systems, where labels are often incomplete, inconsistent, or weakly supervised. We introduce evaluation sovereignty, defined as the degree to which performance metrics are independent of label authority and supervision regime, and propose a multi-track evaluation framework that systematically varies training and evaluation label sources. Using hierarchical multi-label classification on large-scale scientific metadata, we demonstrate that models exhibiting strong performance under operational ("silver") evaluation degrade substantially under independent ("gold") evaluation, particularly for fine-grained classification. For example, Micro-F1 decreases from approximately 0.54 to 0.03. Notably, ranking-based metrics remain above baseline, revealing a divergence between latent model signal and classification validity. These findings suggest that commonly reported performance metrics may reflect alignment with labeling processes rather than true predictive capability. We therefore reconceptualize evaluation validity as a system-level property shaped by label governance and provide a practical methodology for auditing intelligent systems operating under weak supervision.

Retrieval over visually-rich documents, pages that interleave text with figures, tables, and charts, is essential for multimodal retrieval-augmented generation, yet most retrievers still discard the visual channel. The \emph{Multimodal Document Retrieval Challenge}, Track~1 of the MIR Challenge at the first EReL@MIR workshop, co-located with The Web Conference 2025, asks participants to build a \emph{single} retrieval system that handles two complementary regimes: closed-set document page retrieval within long documents from a text query (MMDocIR), and open-domain retrieval of Wikipedia-style passages from an image or image-plus-text query (M2KR). Systems are ranked by the macro-average of mean Recall@$\{1,3,5\}$ over the two tasks. The challenge drew 455 entrants and 586 submissions across 22 teams. This report describes the challenge design, datasets, and evaluation protocol; reports the final standings; and analyses the three winning teams' systems. All three build on decoder-based Multimodal-LLM embedders from the Qwen2-VL family rather than on CLIP-style encoders, and differ chiefly in whether they reach the top through fine-tuned ensembles, training-free multi-route fusion with a strong vision-language re-ranker, or zero-shot late interaction. The training-free system finished within $0.1$ point of the fine-tuned winner.

Despite being home to more than 1300 ethnic groups and 700 indigenous languages, bias in Large Language Models has not been fully studied in Indonesia, thus leaving a critical gap in evaluating representational fairness and localized stereotypes within its uniquely vast, multilingual, and diverse sociocultural landscape. To address this, we introduce IndoBias as a culturally-grounded bias benchmark to assess LLMs bias in Indonesian and three local languages: Javanese, Sundanese, and Makasar. IndoBias features dual perspective evaluation tracks: depth-oriented (with contrastive-pairs) and breadth-oriented (with generation-based), where the latter is grounded in social science frameworks (SPI, O*NET, and WGI). Our results show that existing LLMs -- particularly decoder models -- exhibit strong bias towards prototypical sentences in Indonesian, while local languages suffer higher bias under Ideology and Religion category. We also find that LLMs responses exhibit a non-uniform Stereotype Polarity when prompted with various local entities. Finally, we discover that, in Indonesian, Common Crawl texts introduce more bias during pretraining, compared to human-reviewed article texts (e.g., Wikipedia, News), whereas introducing local languages to pretraining generally increases bias. This work highlights the importance of studying bias in culture-specific context. Warning: This paper contains example data that may be offensive, harmful, or biased.

Referring 3D Gaussian Splatting (R3DGS), which utilizes natural language for 3D object segmentation, has emerged as a crucial capability for embodied AI. However, existing methods typically rely on expensive per-scene manual annotation and per-view pseudo mask generation, which suffer from multi-view inconsistency and poor generalization to varying query specificities. To address this, we present TrackRef3D, a fully automatic pipeline that achieves open-world referring segmentation in 3D Gaussian Splatting (3DGS) without manual annotation by introducing a multi-view consistent track-then-label paradigm that fundamentally decouples object discovery from semantic grounding. Specifically, we propose a Trajectory-Aware Semantic Consensus Module (TSCM) which aggregates cross-view predictions via synonymous clustering and trajectory-aware voting to establish a canonical semantic identity, thereby ensuring multi-view consistency. Furthermore, we employ a visibility-aware description generation strategy to mitigate ambiguity and propose a Hybrid Training Strategy (HTS) that jointly optimizes coarse category semantics and fine-grained referential cues to ensure robustness under varying query specificities using a multi-positive contrastive objective. Extensive experiments on benchmarks demonstrate that TrackRef3D achieves state-of-the-art performance.

This report presents our solution for the WeatherProof Dataset Challenge, namely CVPR 2026 8th UG2+ Challenge Track 2: Semantic Segmentation in Adverse Weather. For the semantic segmentation task under adverse weather conditions, we propose a semi-supervised segmentation pipeline. Our method is trained exclusively on the WeatherProof dataset, without using any additional external data. Specifically, we adopt UniMatch V2 as the baseline model and treat all degraded-weather images as unlabeled data for semi-supervised training, thereby fully exploiting the data distribution provided by the challenge. During inference, we further apply test-time augmentation to improve the robustness and segmentation accuracy of the final predictions. The code is publicly available at: https://github.com/ylb888/weatherproof-challenge-unimatchv2.

Segment Anything Model 2 (SAM 2) serves as a core foundation model in the field of video segmentation. Building upon the original SAM model, it introduces a memory bank mechanism and demonstrates outstanding performance in tasks such as semi-supervised video object segmentation and tracking anything. However, the complex computational characteristics of SAM 2's multi-stage image encoder and memory module have raised the barrier to the model's deployment in practical applications. To address this issue, we propose TinySAM 2, a lightweight video segmentation model that balances performance and efficiency. First, a memory quality management mechanism is introduced to select and retain high-informative historical frames as the memory. In addition, a joint-spatial-temporal token compression is proposed that reduces the memory storage and computational cost. Specifically, average pooling is employed to first compress redundancy tokens in the spatial domain. In the temporal domain, informative tokens are selected across frames in the memory bank based on token-level similarity measurement. Besides, we take RepViT as the lightweight image encoder, which further reduces the model parameters. Extensive experiments on challenging datasets such as DAVIS and SA-V demonstrate that TinySAM 2 achieves 90% of the performance of SAM 2.1, with only 7% memory tokens and 3% training data. This study effectively alleviates the bottlenecks in parameter count, computational load, and deployment costs associated with SAM 2, providing a resource-efficient solution for the widespread application of video segmentation models on devices.

Visual anomaly detection (AD) for industrial inspection is a highly relevant task in modern production environments. The problem becomes particularly challenging when training and deployment data differ due to changes in acquisition conditions during production. In the VAND 4.0 Industrial Track, models must remain robust under distribution shifts such as varying illumination and their performance is assessed on the MVTec AD 2 dataset. To address this setting, we propose a training-free and class-agnostic anomaly detection pipeline based on the work of SuperAD. Our approach improves generalization through several modifications designed to enhance robustness under distribution shifts. These adaptations include using a DINOv3 backbone, overlapping patch-wise processing, intensity-based augmentations, improved memory-bank subsampling for better coverage of the data distribution, and iterative morphological closing for cleaner and more spatially consistent anomaly maps. Unlike methods that rely on class-specific architectures or per-class hyperparameter tuning, our method uses a single architecture and one shared hyperparameter configuration across all object classes. This makes the approach well suited for industrial deployment, where product variants and appearance changes must be handled with minimal adaptation effort. We achieve segmentation F1 scores of $62.61\%$, $57.42\%$, and $54.35\%$ on test public, private, and private mixed of MVTec AD 2 respectively, thereby outperforming SuperAD and other state-of-the-art methods. Code is available at https://github.com/LukasRoom/SuperADD.

Audio-based video object segmentation aims to locate and segment objects in videos conditioned on audio cues, requiring precise understanding of both appearance and motion. Recent audio-driven video segmentation methods extend MLLMs by fusing audio and visual features for end-to-end localization. Despite their promise, these approaches are computationally intensive, struggle with aligning temporal audio cues to dynamic video content, and depend on large paired audio-video datasets. To address these challenges, we present ASR-SaSaSa2VA, a resource-efficient framework for audio-guided video segmentation. The key idea is to convert audio inputs into textual motion descriptions via automatic speech recognition (ASR) models and then leverage pre-trained text-based referring video segmentation models (e.g., SaSaSa2VA) for pixel-level predictions. To further enhance robustness, we incorporate a no-target expression detection module, implemented by a fine-tuned audio-based MLLM, which filters out audio clips that do not refer to any target object. This design allows the system to exploit strong pre-trained models while effectively handling ambiguous or irrelevant audio inputs. Our approach achieves a final score of 80.7 in the 5th PVUW Challenge (MeViS-v2-Audio track), earning the second-place ranking.

We present NightFeats, a structured multi-agent retrieval-augmented generation (RAG) system submitted to the MMU-RAGent competition at NeurIPS 2025, where it was awarded Best Dynamic Evaluation in the text-to-text track. Rather than targeting benchmark maximization, this work proposes a principled pipeline that decomposes knowledge synthesis into three coordinated phases: retrieval, curation, and composition, each governed by explicit intermediate representations and handoff contracts. Inspired by Agentic Context Engineering (ACE), the system introduces temporal-semantic reranking, bounded contradiction reconciliation, and citation-preserving composition as core architectural primitives. Competition results show that NightFeats surpasses proprietary baselines including Claude-SonnetV2 and Nova-Pro on LLM-as-a-Judge and Human Likert evaluations, confirming that architectural transparency and verifiable evidence grounding are better aligned with human preferences than systems optimizing narrowly for automatic similarity metrics.

Full-duplex dialogue audio, in which each speaker is recorded on a separate track, is an important resource for spoken dialogue research, but is difficult to collect at scale. Most in-the-wild two-speaker dialogue is available only as degraded monaural mixtures, making it unsuitable for systems requiring clean speaker-wise signals. We propose DialogueSidon, a model for joint restoration and separation of degraded monaural two-speaker dialogue audio. DialogueSidon combines a variational autoencoder (VAE) operates on the speech self-supervised learning (SSL) model feature, which compresses SSL model features into a compact latent space, with a diffusion-based latent predictor that recovers speaker-wise latent representations from the degraded mixture. Experiments on English, multilingual, and in-the-wild dialogue datasets show that DialogueSidon substantially improves intelligibility and separation quality over a baseline, while also achieving much faster inference.

As people increasingly rely on artificial intelligence (AI) for guidance in their own lives, scholars, lawyers, and even judges have begun to consider the role of AI in legal decision-making. As "silicon sampling" -- the use of generative AI models in social science research -- is now impacting academia, "silicon jurors" could make an appearance in courtrooms. This study joins an emerging line of research on generative AI models' ability to simulate human legal judgments. In particular, we study how large language model (LLM)-powered chatbots respond to series of questions about legal reasonableness. When the law needs to judge the appropriateness of a behavior, it most often asks whether the behavior was "reasonable." Yet despite the ubiquity of reasonableness judgments, they are the site of constant vexation for lawyers, judges, and lay people. Reasonableness seems inherently vague and unpredictable, since it relies on variable context and implicit conceptual schemas. Moreover, many scholars caution that reasonableness judgments may vary along demographic lines. We compare the answers of human participants to those of twenty-six LLMs across twenty-five different legally relevant reasonableness judgments. Overall, our findings suggest that chatbot responses generally track those of human participants. Nonetheless, we find some suggestive -- and potentially concerning -- results. Compared to humans, LLMs generate more homogeneous responses and occasionally treat a variable standard as an invariant rule. And, compared to humans, LLMs tend to generate answers that are more favorable to the government and to corporations. Finally, our results indicate that LLMs' responses tend to align more closely with those of respondents who are white, male, older, and more educated. More systematic research is needed to confirm or reject these initial findings.

Language models advise people, keep them company, and write software while they sleep. Measuring what they do is hard: behavior has to be sampled repeatedly across models, prompts and releases, most of it lives in unstructured text that has to be coded before it can be counted, and the result has to be legible and rigorous enough to meaningfully compare models and vendors. To address these constraints, we present a simple, cheap, scalable, and replicable model for studying model behavior. Each study is a frozen, public stimulus run identically on a cross-vendor panel, at a few dollars per model or less. Each reads its transcripts one of three ways, chosen by how much interpretation the behavior needs: exact match on a clamped reply, a codebook applied by LLM judges whose agreement with a human coder is reported per code, and an instrumented environment that records what an agent did independently of what it said. Run across four years of model releases from both frontier and open-source labs, these instruments find four things. Convergence: asked to pick a word, 27 of 44 models answer serendipity at least once in four tries. Resistance: a trailing "right?" moves endorsement by up to 32 points, and the sign flips from sycophantic to resistant as generations advance, keyed to the tag's surface form. House: whether a model holds a position under pressure tracks its generation, and how it holds tracks the lab that built it. Account: told to do something the documentation in their repository contradicts, some coding agents never went along silently and others always did, and the same model can change with the harness it runs in. Re-run on every release, batteries like these track how behavior is changing across vendors and over time.

ArGuard is a shared task on harmful content detection in Arabic memes and LLM prompts. It includes two tracks: Track A focuses on multimodal hate detection in Arabic memes, while Track B addresses harmful prompt detection for Arabic LLM safety evaluation. In total, 58 teams registered, 35 participated in the final evaluation, and 27 submitted system-description papers. Participating teams explored models such as AraBERT, Jais, and Qwen3-VL. The best systems achieved macro-F1 scores of 0.823 on A1, 0.419 on A2, 0.984 on B1, and 0.790 on B2. Fine-grained meme classification in A2 was the most challenging setting, partly due to sparse labels and train-test distribution shifts.

Outdoor robots require more than an accurate receiver and a path-tracking law: the navigation system must preserve geometric consistency from geographic waypoints to actuator commands, expose measurement validity and timing, and respond to invalid or stale state information. This work presents a ROS2 navigation stack with interchangeable single-GNSS--IMU and dual-antenna-GNSS localization front ends. Both provide a common local East--North--Up state interface for pure pursuit, virtual-point cross-track PID, finite-horizon nonlinear model predictive control (NMPC), and a segment-dependent hybrid dispatcher. The architecture specifies coordinate conventions, datum initialization, asynchronous state construction, waypoint geometry, controller equations, quality gates, command arbitration, and watchdog behavior. Independent physical field runs collected during 2025 and 2026 grape-vineyard deployments support a balanced evaluation of 800 runs, with 100 runs for each of eight controller--localization combinations on an approximately 199.6-m route. The row-hybrid mode yields the lowest run-averaged post-acquisition mean absolute cross-track error (MAE) in the evaluated dataset: 0.00952m with single GNSS+IMU and 0.00846~m with dual GNSS. These findings characterize deviations of the recorded positions from the reference route under the evaluated conditions. The open-source navigation software and deployment instructions are available in the https://github.com/YiyuanLinXX/PPBv2/tree/main/PPBv2_Navigation.

Text-to-motion models generate plausible human motion but do not model a robot's dynamics; whole-body tracking controllers execute robot references reliably but cannot replan an infeasible one. Recent language-to-humanoid systems bridge this gap by training. We measure how much of the gap closes with no training at all, by putting the deployment controller itself in the loop. Sample-simulate-select (S$^3$) draws $N$ motions per prompt from a frozen text-to-motion model, retargets each to a Unitree G1 by direction-matching inverse kinematics, rolls all of them out under full rigid-body dynamics with the pretrained SONIC tracking policy, and keeps the candidate the policy executed best. Because the verifier is the deterministic simulator itself, S$^3$ attains the any-of-$N$ ceiling by construction; what we measure is where that ceiling lies and what falls short of it. On 200 stratified HumanML3D test prompts with $N=8$, upright execution rises from 83.5% to 89.5% and hardware-gate passes from 33 to 85; on the complete test split (4,184 prompts) it rises from 80.5% to 89.5%. A kinematic verifier that predicts falls well (AUROC 0.90) recovers only a quarter of this gain: ranking a prompt's own candidates is harder than classifying the population. What selection cannot fix is one class, prompts that lower the pelvis, which a generator trained on retargeted robot data does execute. We further score the semantic fidelity of the executed motion with the standard text-motion evaluator, with a real-mocap control that attributes the loss to the robot projection, ablate the retargeter against GMR (complementary failures: the any-of-8 ceiling rises to 95.0% over both), and execute all 177 gate-selected clips on the real G1: every one completes standing, with hardware tracking error matching simulation ($r=0.94$).

Linear RNNs based on the delta-rule enable efficient sequence modeling, but their linear updates with a low-rank correction constrain their expressivity. Prior work has shown that composing two delta-rule transitions in a single recurrent update can model a 2D rotation, but this increases the rank and the cost of the updates compared to a single transition. We show that Kimi Delta Attention (KDA) can realize 2D rotations by combining a single delta-rule transformation with a second reflection supplied by its channel-wise gate. This requires extending the parameter ranges of KDA by combining two existing range extensions: allowing gates in $[-1,1]$ and the delta-rule coefficient $β$ in $[0,2]$. We call the resulting model Complex KDA (CKDA). It preserves KDA's stability and efficiency, with transitions that remain diagonal-plus-rank-one and non-expansive, while reaching the state-tracking expressivity of DeltaProduct$_2$. We characterize the expressivity of CKDA and prove that every orthogonal diagonal-plus-rank-one matrix is exactly a CKDA transition matrix. A single CKDA layer can track every finite group isomorphic to a subgroup of $\mathrm{SO}(3)$, and many state-tracking results use one fewer layer for CKDA compared to other diagonal-plus-rank-one Linear RNNs. Empirically, combining both extensions yields the strongest length extrapolation among tested KDA range settings on $S_3$, $S_4$, and periodic audio continuation. In language modeling, CKDA outperforms Transformers and other linear RNNs, obtains similar results to a KDA baseline, and shows promising scaling behavior. Our code is open source at https://github.com/OpenEuroLLM/ComplexKDA and our models are available at https://huggingface.co/collections/openeurollm/complexkda.

Evaluating world models requires assessing both the quality of the worlds they generate and their consistency and responsiveness under exploration, interaction, and modification. We introduce HappyWorld-Bench, a comprehensive benchmark that evaluates whether generated worlds remain reliable as agents interact with them. Our design is built on a hierarchical capability framework of six world capabilities (W1-W6), from generative construction to unified world modeling, instantiated across three independent evaluation tracks: video world models, spatial world models, and embodied world models. HappyWorld-Bench comprises 1,138 video prompts, 300 spatial scenes, and 254 embodied test cases. Across all three tracks, we build and operate HappyWorld-Arena to organize human A/B comparisons and derive model-level Elo ratings, which complement newly designed automated metrics that capture behavioral correctness. We evaluate 14 video world models, 9 spatial systems, and 8 embodied candidates under this unified framework. Results reveal remaining reliability gaps across all three tracks: video models exhibit reduced consistency during extended rollouts and revisits, spatial models achieve at best 70.14% placement accuracy and 73.33% edit execution, and embodied models struggle to preserve state across multi-step actions and respond precisely to altered action conditions and physical rules. These findings highlight the need to evaluate world models not only by visual quality, but also by state consistency and the correctness of their responses to actions and interventions.