video analytics developers

Abstract: Cross-view object correspondence (CVOC) aims to establish associations of the same target between query and search videos captured from different viewpoints. Existing methods often rely on visual cues for CVOC. Despite recent progress, these methods struggle in complicated scenarios with substantial viewpoint changes and appearance variations, due to lacking sufficient target information. Addressing this, we introduce TeSCo, a framework that exploits rich Textual Semantics of an object, in addition to its visual cues, for cross-view object Correspondence. Our key insight is, textual description of a target, by capturing diverse attributes such as category, texture, and color, provides viewpoint- and appearance-invariant semantics, which are complementary to visual cues and can hence enhance correspondence robustness in complicated scenes. Inspired by this, TeSCo first generates a textual expression of the target from the given query view using a vision-language model, and then extracts textual feature to enhance visual features for localization in the search view. To realize this, we introduce a text-conditioned cross-modal fusion (TCF) module that incorporates the textual feature into visual features with text-guided modulation, producing more robust multimodal target representation for object correspondence. Since not all textual cues are equally compatible with the target in current search-view frame, we propose an iterative textual feature refinement (ITFR) module that progressively adjusts textual feature using search-view information before the TCF module, enabling search-view-aware textual semantics to be fused into visual features and leading to better performance. In extensive experiments on Ego-Exo4D and HANDAL-X, TeSCo demonstrates state-of-the-art results and largely outperforms other models, validating its efficacy. Our code is released at https://github.com/VAIL-Fan/TeSCo.

Abstract: Streaming video question answering (QA) requires real-time reasoning over continuous video streams under strict latency and active-context budgets. Current streaming vision-language models (VLMs) often give confident yet ungrounded answers when queries are ambiguous or have false premises, or when relevant visual evidence is no longer in the active context. Proactive clarification can help, but must satisfy the same latency and memory constraints. We present StreamClarify, a training-free framework for grounded streaming QA that combines a pre-answer uncertainty check with hierarchical memory. The check uses the entropy of a binary pre-answer judgment as a lightweight uncertainty score to decide whether to answer directly or request clarification. To support re-answering, a temporal-pyramid KV cache maintains a bounded active visual context, while a CPU-side semantic retrieval archive provides access to historical frames. We further construct StreamClarify-QA, a diagnostic benchmark of 1,837 QA pairs from StreamBench and OVO-Bench, spanning standard, referential ambiguity, and unanswerable scenarios. Across three VLM backbones, StreamClarify improves Macro accuracy by up to 8.2 points over backbone-matched baselines. At comparable trigger rates, entropy-based selection achieves more than 18 points higher precision than random selection in identifying initially incorrect answers. Under oracle user guidance, clarification improves Macro accuracy by up to 3.5 points. We also report a controlled, oracle-assisted Qwen3-VL simulation on Ambiguous QA with generated questions and referent-only user replies, where the entropy-gated condition improves accuracy by 8.3 points.

Abstract: Dropwise condensation is key to power and water systems. State-of-the-art condensation surfaces are often inhomogeneous, either patterned by design or random defects. Inhomogeneous surface features dictate the rates of each stage of dropwise condensation-nucleation, growth, coalescence, and shedding. Characterizing these rates requires the precise detection of thousands of droplets. Yet, existing segmentation methods are unable to detect droplets on inhomogeneous surfaces. We developed TileSAM3, a universal droplet segmentation tool built on the Segment Anything Model 3 (SAM3). Hierarchical tiling with per-tile inference followed by post-processing extended SAM3's detection cap and raised the effective resolution of small droplets. We created two datasets: High Quality Droplet Dataset (N=365,826; HQDD) with homogeneous surfaces and Low Quality Droplet Dataset (N=63,577; LQDD) with patterned, defect-laden surfaces under adverse lighting. TileSAM3 closely reproduced thresholding annotations on HQDD, surpassed it over sixfold on LQDD (Average Precision 0.57 versus 0.088), and outperformed other deep learning models on all metrics evaluated. Combined with KD-tree tracking, TileSAM3 identified localized nucleation and coalescence in narrow active bands on a stiffness-patterned surface. TileSAM3 enables analysis of understudied inhomogeneous substrates with poor lighting and videography which facilitates faster, lower cost research of condensation materials.

Abstract: Cross-view object correspondence identifies the same object across views. However, existing methods often rely on frame-wise matching or sequential tracking, which can propagate correspondence errors under appearance changes or occlusions. Therefore, we introduce cross-view temporal object correspondence, which requires preserving object identity across both views and time from a query mask. To address this, we propose XTrack, a unified framework which repurposes a pretrained video DiT with semantic and spatiotemporal priors for RGB-to-mask latent transport, jointly modeling cross-view matching and temporal tracking. However, naive video DiTs do not guarantee accurate cross-view grounding and often attend to semantically similar yet incorrect objects across views. Thus, we introduce focal cross-view attention alignment, hard negative conditioning, and local context crop strategies to ensure reliable grounding. Finally, beyond these architectural enhancements, we also tackle the lack of dense annotation by introducing a pseudo-label curation pipeline that converts sparse Ego-Exo4D annotations into dense supervision to enable effective video-level training. Our results demonstrate the effectiveness of XTrack as a unified framework, ensuring robust temporal consistency and identity preservation.

Abstract: Chain-of-thought reasoning has substantially advanced video understanding in multimodal large language models (MLLMs). However, language-based reasoning traces do not provide a reliable mechanism for models to revisit relevant visual evidence, leading to visual forgetting and impairing their ability to reason over video content. Existing approaches maintain visual grounding during reasoning in two main ways: they either refer to evidence using textual coordinates, which specify symbolic locations rather than the visual content itself, or reinsert cropped regions or frames into the multimodal context, preserving visual details at the cost of substantial context expansion. To address these limitations, we introduce VideoLVR, which constructs latent visual evidence from local regions at relevant video moments, thereby providing visual support for intermediate reasoning steps. To supervise latent visual evidence generation, we introduce a visual token prediction objective that formulates visual token reconstruction as prediction over a vocabulary derived from the input video, thereby unifying visual evidence reconstruction and text generation within the same token prediction framework. This shared prediction framework enables MLLMs to reason over interleaved text tokens and latent visual states while integrating evidence from different moments. We further extend VideoLVR to reinforcement learning by treating sampled visual tokens as actions alongside text tokens and jointly optimizing the policy using final-answer rewards. By efficiently revisiting task-relevant visual information during reasoning, VideoLVR improves video understanding without substantially expanding the multimodal context. Experiments on multiple video benchmarks demonstrate its effectiveness, especially on tasks requiring fine-grained visual evidence.

Abstract: Event-triggered recording needs a useful combination of target capture and false triggers, yet representation comparisons can change with readout and threshold calibration. We evaluate real event accumulations from 520 DailyDVS-200 clips and 47 participants using participant-disjoint training, validation, and testing. The original study varies frozen encoder scale, accumulation duration, and temporal view count; a post-review extension adds two supervised-pretrained convolutional encoders, a 61,586-parameter trained CNN with three seeds, ten repeated participant partitions, and calibration sensitivity. At 50 ms/eight views under the primary handshake-or-fall policy, DINO-S CLS captures 101/132 targets with 40/388 false triggers, versus 78/132 and 35/388 for patch mean. DINO-B CLS has lower selected-threshold capture but higher fold-wise low-FPR partial AUC (0.746 versus 0.732). MobileNetV3-Small captures 97/132 with 53/388 false triggers, demonstrating why lightweight baselines and actual operating points matter. Across ten dependent repartitions, mean normalized partial AUC is 0.738 for DINO-B CLS and 0.674 for MobileNet. The original LeJEPA S-to-B capture difference is 8.3 points, but its multiplicity-adjusted interval includes zero. Duration and view-count effects remain policy dependent. We find no established JEPA-specific advantage. The evidence concerns offline action-defined candidate selection on previously used development data; it does not establish personal recording value, continuous false alarms per hour, or eyewear energy.

Abstract: An action seen once and the same action seen three times receive the same action-recognition label, so that supervision does not require a model to count. Answers shared across histories that differ in event order, accumulation, or relations leave those distinctions unconstrained; we call this state under-specification. Visible-State Supervision (ViSS) addresses it by training on source-grounded probes of event identity, order, accumulation, relation change, and answer sufficiency; under the same recent-frame interface it improves OVO-Bench over a strict-matched generic temporal supervision baseline while remaining comparable on StreamingBench, with the largest gain in hallucination detection. Test-Time Causal View Search (TVS) then adapts evidence access while keeping the learned model fixed, expanding from recent frames to sparse event and compact history views on demand. The two interventions separate representation from view selection: supervision particularly benefits hallucination detection and causal reasoning, access particularly benefits event recall and forward reasoning. Together they reach the best reported accuracy among the published systems compared here on StreamingBench and OVO-Bench at a fraction of the processed frames of exhaustive selection, and strict replay shows that most of the remaining gap to a three-view oracle is a correct answer that was later replaced. Streaming accuracy depends on what supervision asks the model to keep, what evidence inference exposes, and whether the model keeps a correct answer once it has one.

Abstract: Recently, point-supervised temporal action localization has attracted increasing attention due to its low annotation cost (only one frame per action instance) and competitive detection performance. However, since point annotations are less complete than proposal annotations, existing methods mainly rely on multiple instance learning for snippet-level classification, which leads to a performance gap compared to fully supervised models. To bridge this gap, we propose P2P-Former, a novel point-to-proposal Transformer-based framework that effectively connects point-level supervision with proposal-level learning for temporal action localization. In this work, we identify two key challenges in transitioning from point-level supervision to full supervision: (1) how to generate reliable proposals as pseudo ground-truth, and (2) how to effectively integrate point-level cues into the fully supervised paradigm. To address the first challenge, we propose a Point-aware Proposal Generation (PPG) strategy that refines the confidence of proposals produced by a base locator using a point-aware importance estimation function, and then performs optimal matching between point annotations and candidate proposals to select reliable proposals as pseudo ground-truth. To tackle the second challenge, we introduce a Point-guided Semantic Enhancement (PSE) module that aggregates prototypes from point-annotated features and injects them into snippet features through Transformer-based interactions. Moreover, we design a Point-to-Proposal Constraint Loss (PPCL) that enforces each predicted proposal to include the annotated point, thereby encouraging more accurate localization. Extensive experiments on the THUMOS14 and ActivityNet-1.3 datasets demonstrate that our method outperforms several existing state-of-the-art approaches.

Abstract: Camera motion recognition requires separating viewpoint changes caused by camera translation, rotation, or zoom from object motion and scene dynamics. Video-language models often rely on semantic cues and struggle with the geometric evidence needed for this task. Joint-Embedding Predictive Architectures such as V-JEPA offer strong spatio-temporal representations, but our experiments suggest that they underrepresent geometric cues in some regions. Under temporal-reversal supervision, V-JEPA improves overall but degrades on several motion families, whereas a weaker geometry-oriented VGGT predictor improves across all families. We refer to these geometrically underrepresented regions as geometric blind spots and propose Fovea-JEPA to identify and address them. Using V-JEPA as the base model and VGGT as a geometry expert, we construct intervention pseudo-labels from their prediction errors to weakly supervise a blind-spot predictor. The predictor produces token-wise response maps, which are aggregated into a clip-level intervention score to guide adaptive fusion of visual and geometric features. This selective intervention introduces complementary geometric information while preserving reliable visual predictions. Experiments on CameraMotionDataset, CameraBench, and ShotBench demonstrate the effectiveness of Fovea-JEPA, with gains of 9.67, 8.67, and 24.38 percentage points over previously reported state-of-the-art baselines, respectively. Code and models will be released upon acceptance.

Abstract: Attribute-based person retrieval ranks surveillance images from a binary query of semantic attributes, but mean Average Attribute Degree of Match (mADM) drops on unseen cameras as predicted probabilities shift. Common pipelines rank gallery images by their L1 distance to the query, which penalizes a confident single-attribute mismatch no more than small uncertainties across many attributes, while post-hoc recalibration fitted on source validation data fails to transfer because the per-attribute maps fitted on source data do not match those of new cameras. We propose LENS (Likelihood Evaluation and Neighbor Smoothing), combining fine-tuned DINOv3 and CLIP models with a distilled student. LENS regularizes mutually exclusive attribute groups with group cross-entropy in training, scores gallery images by the log-likelihood of the exact query, and smooths gallery logits over nearest neighbors using DINOv3 appearance features from the same forward pass. On held-out datasets, all three changes raise mADM: the likelihood score improves all six retrieval metrics for the main ensemble and mADM for all 48 development models, and training-free kNN smoothing improves all 46 tested development models. On the hidden test set of the UPAR 2027 benchmark, LENS reaches 68.17 mADM.

Abstract: Video reasoning requires understanding how events unfold over time, yet videos with accurate temporal annotations are expensive to collect. Self-play offers an alternative by letting a model generate and solve its own tasks without manual annotation. However, existing video self-play methods train on fixed sets of recorded videos, so they cannot alter the underlying event dynamics and rely on noisy pseudo-labels derived from the model's own outputs, such as majority voting, that risk reinforcing shared model errors. To overcome these limitations, we introduce Tempo-Zero, a self-improvement framework that closes a loop of executable environments, execution-verified self-play, and environment evolution, where executing a program renders a video with the ground-truth option and interval of every question. A single policy plays both Questioner and Solver, proposing questions and answering them with timestamped evidence lines, and is trained on execution-grounded rewards instead of model consensus. The Solver's performance steers a curriculum over tasks and difficulty levels, and the policy selects rule changes that enrich the event dynamics of later rounds. Despite being trained solely on simple programs, without supervised fine-tuning or external video data which prior methods typically require, Tempo-Zero outperforms the base models and recent self-play baselines on average across twelve real-world video benchmarks at different model scales, with particularly strong gains on temporal grounding tasks. Project page: https://alchemist-jo.github.io/tempo-zero/

Abstract: Humans efficiently process multimodal streams spanning hours to years, building persistent, fine-grained spatiotemporal memories that support reasoning about and predicting long-horizon events. In contrast, existing benchmarks focus on short- and medium-context tasks, and most foundation models struggle to preserve fine-grained information over long horizons. To evaluate and improve long-horizon spatiotemporal reasoning, we introduce the LongVideo dataset and the 4DMemoryBench. LongVideo comprises about 900 hours of video and over 5,000 manually reviewed questions covering visual, textual, and spatiotemporal reasoning across diverse context lengths. We further develop the 4DMemory Harness, which integrates four complementary skills, 22+ specialized tool models, and a model-specific retrieval-augmented generation system. Evaluating 13 leading models, we find that standalone foundation models struggle with long-horizon tasks, while the 4DMemory Harness improves all of them by over 50\%. Compared to previous benchmark, the same models exhibit a performance drop of more than 10\% on 4DMemoryBench. We hope 4DMemory will facilitate the development of spatiotemporal agents capable of addressing complex, long-horizon tasks.

Abstract: Vision language models (VLMs) have been increasingly adopted in consumer smart-home camera systems to interpret recorded videos and provide insights. However, video clips captured from smart-home cameras are different from curated web videos dominating existing VLM benchmarks. Smart-home video clips often do not contain subtitles or narration and are captured from different viewing angles, under various lighting conditions, and in diverse scenarios. However, the capabilities of the model in this domain, under sparse visual signals and limited context, are not fully assessed by existing VLM benchmarks. To close this gap, we introduce OpenHome, a visual-only, four-option multiple-choice benchmark for evaluating smart-home video understanding. It contains 2,254 questions on 1,118 real-world video clips and organizes 10 tasks into three capability levels. Perception evaluates visible content, Reasoning evaluates causal, spatial, and temporal relationships, and Insight evaluates subject intent, supporting evidence and appropriate homeowner responses. Based on the dataset, we describe the task design, data pipeline and evaluation protocol, and report results for 32 closed-source and open-weight VLMs. The best model, Gemini 4 Argon, reaches 81.2% overall accuracy. Frontier closed-source models outperform open-weight models on every level. The best closed-source model leads the best open-weight model by 16.3% on Perception, 22.4% on Reasoning and 37.4% on Insight.

Abstract: Joint-embedding predictive architectures (JEPAs) learn world models in latent space, enabling efficient planning but leaving their states opaque. We introduce Inside-JEPA, which places relational channels for ordered entity pairs inside the JEPA latent state and shows that an inverse-dynamics loss can ground these channels without relation labels when its head reads them, either alone or as part of the full state. On Something-Something V2, latent prediction alone performs no better than a random relation head, while inverse-dynamics supervision that recovers actions from channel changes improves relational AUROC by .065-.103 across ten settings with annotated boxes, by .081 on RH20T robot episodes with held-out tasks, and by .050-.061 across five relations using off-the-shelf detection and tracking. An inverse head on entity latents decodes actions more accurately but barely improves relational grounding; shuffled labels, two label-free objectives, and action labels on only 5% of the pairs also fail to ground the channels. Given an action, predicted relational changes approach the quality of observed future relations (AUROC .599 vs .621). For one-step relational goals, a supervised readout of the 16 channels ranks the correct action among the top five of 65 in 43% of cases, compared with 24% for prediction-only latents and 17% for a frequency prior.

Abstract: Fine-grained multimodal intelligence requires models to both understand a designated region and localize it from language. Existing methods learn these directions with separate objectives and task-specific annotations. We present Generalized CycleGRPO, the first unified reinforcement learning framework coupling region understanding and localization across spatial, temporal, and spatio-temporal domains. Given a region, a multimodal large language model generates candidate descriptions and then localizes each description to reconstruct the input region. Reconstruction overlap provides a verifiable reward for captioning, while individual localization rewards improve the inverse mapping. A common region-space formulation extends this cycle from image masks and boxes to video intervals and timestamped spatial evidence, without reference captions or an external language-model judge during reinforcement learning. In videos, the cycle encourages descriptions that distinguish a target event from similar recurrences. We introduce DOVE (Distinctive Occurrence-level Video-caption Evaluation), a human-verified contrastive benchmark for this temporal distinctiveness. CycleGRPO improves SAMTok's overall gIoU from 57.5% to 68.2% on GroundingSuite and average gIoU from 71.2% to 78.7% on GRES. For temporal tasks, it achieves 51.9% tIoU on Charades and 33.3% Temporal Evidence Coverage on DOVE, compared with 45.6% and 14.8% for the Qwen3-VL initialization. These results establish region-text cycle consistency as a general self-evaluation principle for multimodal understanding and localization. Models, code, and DOVE are available at https://devinxzhang.github.io/Generalized-CycleGRPO-Page/.

Abstract: Streaming video understanding requires a model to decide not only what to answer but also when to answer. Evidence arrives progressively, and recent benchmarks identify premature responses, issued before the queried event has concluded, as a recurring failure mode. Existing approaches often learn response timing through additional training, while training-free alternatives typically rely on scalar gating signals that do not explicitly represent when an answer becomes final. We organize chunk-wise streaming errors into three patterns: sampling variation can make an answer unstable, a stable answer may lack support in the observed video, or an answer may be supported by the current evidence while the event remains unfinished. We propose PACE (Plan, Attend, Check, and Emit), a training-free framework designed around these three failure modes. Given a question, PACE drafts a plan specifying what to observe and what will signal completion. As the stream unfolds, it invokes an off-the-shelf video LLM at informative moments, evaluates each draft for stability, grounding in its cited frames, and completion, and records verified observations in an event ledger. Experiments on ProactiveVideoQA and OVO-Bench Forward show that PACE improves over the official protocols by up to 5.9 average PAUC points and 6.2 average accuracy points across proprietary and open-source backbones, yielding state-of-the-art performance.

Abstract: Class unlearning for video segmentation aims to ensure that a removed category cannot be reconstructed from a deployed model's outputs. Current leakage evaluation algorithms inspect one frame at a time, and we show that this understates leakage by a wide margin: an attacker that averages a model's output field over a two-second clip recovers the forgotten class from every method we evaluate on VSPW and most on Virtual KITTI~2, including models whose per-frame predictions are visually indistinguishable from a retrain oracle (a model trained from scratch without the forgotten class). We attribute the recoverable signal to cross-frame redundancy rather than object trajectory, establish this with a time-shuffled control that leaves leakage intact, and introduce PACE, an unlearning method that suppresses redundancy directly using an exactly permutation-invariant pooled adversarial discriminator and a two-sided temporal consistency objective. Across VSPW and Virtual KITTI~2, PACE lowers pooled attacker AP to $18.79$ (\texttt{person}) and $12.10$ (\texttt{car}) on VSPW and $9.11$ (\texttt{car}) on Virtual KITTI~2, achieving on the latter the strongest suppression among verdict-safe methods retaining usable utility, while retaining $19.8$ mIoU on VSPW \texttt{car} against at most $8.3$ for any baseline achieving comparable erasure.

Abstract: Reinforcement learning with verifiable rewards, particularly Group Relative Policy Optimization (GRPO), has improved video reasoning in multimodal large language models. However, standard video GRPO varies only linguistic trajectories while conditioning every rollout in a group on the same sampled frames, which can produce homogeneous rewards and weaken group-relative learning when the shared visual evidence is insufficient or distracting. We propose \textbf{FrameGRPO}, a simple and effective reinforcement-learning framework that requires neither auxiliary networks nor changes to the model architecture. Specifically, FrameGRPO combines two lightweight components: (1) \textbf{Multi-Frame Rollouts}, which use only uniform temporal subsampling of a single maximal-frame decode to construct aligned views that diversify visual evidence and reasoning trajectories; and (2) \textbf{Gated Temporal Efficiency Credit}, which selectively reinforces a correct lower-ratio trajectory when all rollouts in a fixed full-ratio reference subset fail, without imposing an unconditional preference for fewer frames. Extensive experiments across diverse video understanding and reasoning benchmarks demonstrate robust overall improvements over standard GRPO, highlighting temporal-view diversification as an effective mechanism for exploration and credit assignment in video MLLMs.

Abstract: Human memories are organized in a way that connects events with their corresponding spatial contexts. An AI assistant should also be able to memorize and recall the nexus of events and spaces. Thus we introduce SENRI, a task that studies this space--event connection in long, continuous real-life egocentric video. It contains six task types in two directions: Event$\rightarrow$Space, which recovers spatial knowledge from event cues, and Space$\rightarrow$Event, which retrieves events from regions and trajectories. To evaluate this, we build SENRI-Bench, containing 500 manually annotated multiple-choice questions over 10 uninterrupted self-recorded videos with around 25.3 hours of content. Also, we evaluate three typical long video understanding paradigms: direct perception, evidence seeking, and persistent memory. The task remains challenging across all three paradigms, highlighting substantial room for improvement. Then we show two representative failure modes of existing memory methods and address these evidence gaps with SENRI-Agent, a controller that uses Target-Guided Visual Check (TGVC) to propose local visual check tasks for parallel verifiers and integrate the reports into a working memory to help with reasoning. Built upon existing memory systems, SENRI-Agent improves the performance of existing baselines without modifying their stored memories. We hope SENRI will provide valuable insights into how real-life egocentric understanding should be evaluated, reveal the gaps in existing approaches, and offer a possible direction for improvement.

Abstract: Understanding when during a video a model makes its critical observations is essential for interpreting video-understanding systems. However, existing attribution methods often spread evidence diffusely across near-duplicate frames. We introduce Sufficient Frame Subsets: compact sets of frames in which all other frames are masked, yet each subset independently preserves the model's prediction. We evaluate our method on two complementary benchmarks, UCF101 and Something-Something v2, using both CNN- and transformer-based architectures. Our results show that the proposed algorithm identifies highly compact and effective frame subsets for explaining model decisions. We also find that models capable of relying on a greater number of sufficient frame subsets tend to perform better, possibly because they are more robust, consistent with prior research. Finally, the extracted frame subsets help us analyze model failure cases, reveal which aspects of a video the model has learned to emphasize, and identify opportunities for improving future training procedures.

Abstract: We show that a video can be represented with just a few keyframes without sacrificing a given model's belief about the video. Our method is the first of its kind that extracts these keyframes conditioned on the model's prediction, which is massively helpful for building explainable video models. We then fairly distribute the importance of the video to the model among the keyframes with game theoretically fair Shapley values. Through qualitative and quantitative experiments, we show that keyframe attribution is superior to the traditional frame attribution. Further, we conduct preliminary exploration on highlighting the important changes among the frames by looking at the differences between adjacent keyframes.

Abstract: Temporal action detection in long untrimmed videos suffers from inaccurate localization and low-quality proposals, especially with sparse annotations and ambiguous boundaries. For open-ended video understanding and LLM-driven event analysis, reliable proposals connect temporal localization, event annotation, and semantic understanding. However, conventional temporal action detection (TAD) pipelines lack explicit modeling of proposal localization quality, cross-representation support, and inter-proposal relations. We propose Hybrid Rank-Calibration of Action Proposals (HyrCap), a proposal-centric query-based TAD framework for temporal event understanding. Hybrid combines complementary video features with different localization cues, while Rank-Calibration adjusts proposal ranking confidence to reflect localization reliability and candidate competition. HyrCap treats dense query-based predictions as an over-complete temporal event hypothesis space. Cross-Representation Consensus Calibration, Localization Quality Estimation, and Proposal Redundancy Suppression improve proposal ranking reliability, localization quality, and candidate discrimination. Inspired by LLMs' capabilities in fine-grained semantic recognition, event description, and reasoning, we introduce HyrCap for Semantic Action Understanding, using calibrated proposals for standard TAD and as inputs to LLM-based semantic understanding. We define a staged evaluation protocol to separately measure event localization, segment-level semantic understanding, and combined event understanding performance. Experiments on public benchmarks show that HyrCap achieves state-of-the-art performance across multiple metrics and provides a finer-grained paradigm and evaluation protocol for query-based event segmentation and annotation.

Abstract: Thermal-to-visible (T2V) face translation has important applications in nighttime biometric recognition, low-light surveillance, and human interpretation of thermal observations. The task remains challenging because thermal images do not directly record visible appearance, while the cross-spectral cues useful for reconstruction vary across frequency components and semantic roles. Existing spatially conditioned models tend to process these cues through a homogeneous adaptation path and do not explicitly separate local structure from global identity. We propose FreqID-Control, a parameter-efficient framework that organizes thermal guidance according to spectral reliability and semantic role while preserving the pretrained generative prior. It combines frequency-routed low-rank adaptation, which allocates separate transformation subspaces to low-frequency structure and high-frequency detail, with shared cross-spectral identity conditioning, which supplies compact global identity tokens independently of spatial thermal tokens. A controlled analysis on MCXFace and SpeakingFaces provides the empirical basis for this design, revealing stable low-frequency correspondence and edge-localized detail correspondence. Experiments on both datasets demonstrate strong image fidelity and biometric identity preservation, while controlled ablations on MCXFace confirm the complementary contributions of frequency routing and shared identity conditioning.

Abstract: Test-time agentic computation can improve multimodal large language models (MLLMs) through planning, tool use, iterative perception, and reflection, but indiscriminate computation may duplicate resolved work, introduce noisy evidence, and perturb correct predictions. We propose that such computation should be residual: allocate additional computation only to requirements unresolved by the base model and, after a localized failure, recompute only the affected states. We instantiate this formulation in ReTAC, a training-free multiagent framework for video spatial reasoning. ReTAC organizes supplementary computation around unresolved requirements. It plans ordered steps to acquire missing spatial evidence and combines the observations with compatible native evidence for reasoning. In addition, it uses stage-aligned verification to localize failures to planning, perception, or reasoning. Dependency-based recovery recomputes only affected states while reusing unaffected ones. Extensive experiments show that ReTAC achieves the highest overall score among the compared methods across three mainstream MLLMs and two widely adopted benchmarks. Further analyses show that gains from additional computation depend not only on how much is used, but also on where it is allocated and which valid intermediate states are preserved.

Abstract: Not every moment in a video is equally useful for recognizing an action. A few decisive moments often carry the evidence: the instant an object moves, the peak of a dive, or one object placed on another. Existing video datasets rarely provide large-scale, action-conditioned labels for which short segments carry that evidence. This work asks whether vision-language models can identify meaningful moments at scale and whether those labels improve action recognition.We introduce Meaningful Moments (MM), a corpus of about 4.58 million VLM-generated per-segment importance scores for nearly 500,000 class-labeled videos spanning Something-Something v2 (SSv2), Kinetics-400 (K400), and Diving-48 across 622 action classes. We freeze the action classifier and vary only temporal sampling. An alpha-parameterized comparison of importance-led selection against an importance-inverted control separates gains from the VLM signal from gains caused by variable-density sampling alone. Across all three datasets, VLM-derived importance is informative: at the hard-cut endpoint, keeping VLM-important segments outperforms keeping VLM-designated filler. Usefulness depends on class structure. On Diving-48, importance-led fast-forwarding preserves full-video recognition while the inverted control collapses by roughly 30 percentage points. On K400, hard-cut selection beats uniform sampling by 1.45 percentage points, rising to 2.05 under threshold selection. On SSv2, the labels carry signal but do not recover full-video performance. Direct scoring matches greedy removal within its noise floor at roughly 25 times fewer oracle calls, and a four-VLM cross-oracle study shows consistent downstream trends. We release the importance labels and evaluation protocol under CC-BY-4.0; SSv2 class strings remain under CC-BY-NC-4.0.

Abstract: Recent trackers built on foundation models such as the Segment Anything Model (SAM) achieve strong performance by propagating object masks through memory. However, this strength is also a key source of failure: under occlusion, reappearance, or distractors, an incorrect prediction can enter the tracking loop and be propagated over time, resulting in persistent identity drift. Existing approaches seek to prevent such failures through improved memory and temporal modeling, but once a failure occurs, the tracker has no explicit mechanism to recognize or correct it. We address this gap with Dual-Agent, a failure-aware tracking framework that complements propagation with explicit verification and recovery. We address this gap with Dual-Agent, a failure-aware tracking framework that complements propagation with explicit verification and recovery. A vision-language verification agent first reasons over spatio-temporal context to determine whether the tracker is still following the target. When a failure is detected, a correction agent uses temporal context to recover the target before tracking continues. The correction agent is trained on real SAM3 failures, augmented with targeted identity-switch and distractor perturbations that expose it to challenging ambiguities during recovery. On LVOS-v2 and SecVOS, Dual-Agent improves SAM3 by 1.2 and 2.7 $J\&F$, respectively. Importantly, the proposed framework transfers without retraining to SAM2.1, SeC, and Cutie, as well as to long-term video object tracking benchmarks including LaSOT-ext, TNL2K, and VOT-LT2022. Our results show that explicitly reasoning about when tracking fails and how to recover provides a complementary path to robust tracking beyond improving propagation alone. Project page: valaybundele.github.io/Dual-Agent.

Abstract: Video gaze prediction is led by gaze-trained models, yet gaze-free priors carry signal those models have not absorbed, if one knows when to trust them. We propose FocusGate, a gated ensemble of gaze-free priors whose members may abstain. A per-frame gate reads three shape statistics of a defocus map and selects the frames on which the estimator is above chance on average, so rejected frames reduce to the base exactly, while midrank normalisation lets an all-zero prior abstain at zero parameters. Gated fusion is significantly positive on film, sports and web video, whereas unconditional fusion is harmful on sports and null on web. Added to four supervised predictors, the NTIRE 2026 champion among them, FocusGate improves all sixteen model-domain cells in shuffled AUC, fifteen significantly, one domain pre-registered and scored once, while adding only 1% to the champion's latency. Alone, it surpasses TASED-Net and UNISAL in shuffled AUC on film with a 16-frame causal mean.

Abstract: Modern action recognition models rely on video transformers pretrained on massive collections of web-crawled videos, such as Kinetics-700. However, the use of such data raises ethical concerns, as subjects' consent is typically not obtained. Recent high-quality synthetic video datasets generated from motion-capture data, such as BEDLAM2.0, offer a promising ethical alternative. In this work, we investigate self-supervised pretraining of video transformers on synthetic human-motion datasets. We first show that directly applying the standard VideoMAE masking strategy leads to substantially worse performance than pretraining on Kinetics. To address this limitation, we propose a human-centric masking scheme that leverages body keypoints and person bounding box regions. Our approach encourages the model to focus on the structure and dynamics of human motion during pretraining. Experiments on NTU RGB+D and Toyota-Smarthome demonstrate that our method significantly outperforms standard VideoMAE pretraining on synthetic data, closing 49% of the gap to Kinetics pretraining on NTU RGB+D cross-view-subject without using a single real frame during pretraining. To promote the use of ethical action recognition models, we will publicly release our pretrained models.

Abstract: Long-video understanding places substantial demands on memory, as answering questions often requires retrieving information distributed across extended temporal spans. Existing approaches broadly follow two paradigms: query-driven exploration, which is sensitive to localization errors, and query-independent memory construction, which may omit question-specific details. We introduce VideoTapestry, a training-free multi-agent framework that adapts a preconstructed hierarchical video memory through coarse-to-fine, query-driven refinement. The preconstructed memory organizes video content into three levels, capturing global narrative context, event-level temporal structure, and fine-grained relational evidence, respectively. To support coarse-to-fine localization and observation, we assign a specialized agent to each level, keeping retrieval and refinement within a scale-specific context. Guided by the query, these agents revisit relevant video regions and enrich layer-wise memories with targeted multimodal observations. Their refinements are assembled according to the original hierarchy into a composite query-adaptive memory, preserving global context in a compact form while retaining fine-grained evidence along query-relevant branches for final reasoning. Compared with direct GPT-5.5 inference, VideoTapestry achieves absolute accuracy gains of 17.2%, 14.9%, 9.8%, and 7.0% on LVBench, LongVideoBench (Long), Video-MME (Long), and EgoSchema, respectively, achieving the state-of-the-art results among all competitors.

Abstract: Multimodal Large Language Models (MLLMs) have demonstrated remarkable potential in video understanding, yet their reliance on retrospective summarization and text-centric priors often limits their ability to bridge unobserved causal transitions when applied to Video Event Prediction (VEP). To address this, we propose VepAgent, an agentic framework that integrates causal-transition reasoning with tool-augmented reinforcement learning (RL) for robust VEP. Unlike prior methods that passively project future trajectories from historical dependencies, our approach explicitly models the logical progression from terminal observed states to future events. Specifically, we first construct futurebench-4K, a high-quality chain-of-thought dataset for supervised fine-tuning (SFT) that effectively bridges the causal-logic gap by structuring the deduction of unobserved intermediate states. Subsequently, we develop a diagnostic tool library integrating state tracking, frame retrieval, and region magnification, enabling the agent to dynamically augment reasoning with external tools to recover missing spatio-temporal evidence and resolve visual ambiguities during inference. Moreover, we propose a composite reward mechanism that jointly optimizes prediction accuracy, causal coherence, and reliable prior, compelling the agent to rely on genuine visual grounding rather than superficial textual similarities. Extensive evaluations on FutureBench and NEPBench datasets demonstrate that our method achieves state-of-the-art performance, significantly outperforming larger MLLMs and validating the empirical effectiveness of our agentic, future-oriented reasoning paradigm.

Abstract: Multi-instance partial-label learning (MIPL) addresses inexact supervision in both the instance and label spaces, which can be applied to video classification. However, bag-level labels do not explicitly supervise the correspondence between candidate classes and temporal evidence. We propose {\ours}, which couples label disambiguation with temporal evidence allocation through a joint class--time assignment. Occupancy-regularized spherical matching associates contextualized video features while learning nonuniform temporal mass and discouraging excessive concentration. During training, candidate-restricted inference recomputes the assignment within the candidate label set. A dual-marginal KL projection then constructs a structured teacher that incorporates momentum-refined class beliefs while preserving the proposal's temporal occupancy. A single plan-level KL objective aligns the full-space predictor with this teacher. Our analysis characterizes when candidate re-solving differs from masking and shows that, under the stated construction, the joint objective decomposes into class-marginal and class-conditional temporal supervision. We construct VCMIPL benchmarks from Breakfast, DoTA, and FineAction using model-generated candidate labels and evaluate the method across four feature representations. Extensive experimental results demonstrate that PIVOTMIPL outperforms existing MIPL algorithms in both effectiveness and efficiency.

Abstract: Spatio-temporal video grounding (STVG) aims to localize objects or events described by natural language queries in both space and time. Existing STVG models are typically trained and evaluated under the assumption that each query is relevant to the input video. In this work, we challenge this assumption by studying the behavior of state-of-the-art STVG models under irrelevant queries and missing textual input. Our experiments show that current models can still produce plausible spatio-temporal predictions even when the query is unrelated to the video or removed entirely. We further analyze HCSTVG-v2 and VidSTG to identify dataset regularities that may encourage such query-insensitive behavior. Our study highlights an underexplored limitation of STVG models and motivates negative-aware evaluation protocols and architectures that explicitly assess query relevance.

Abstract: Existing streaming vision-language models (VLMs) continuously perceive and reason over visual streams, but their computational pathways remain fixed throughout inference. Consequently, they cannot adapt computation to evolving scene dynamics, where different future events demand different levels and forms of perception. We show that streaming VLMs inherently possess the ability to anticipate the immediate future, and leverage this capability to dynamically configure future computation in a training-free manner. Realizing such anticipatory computation, however, is very challenging: future anticipation must be sufficiently reliable to guide computation, planning must run concurrently with streaming inference, and online reconfiguration must incur negligible overhead. To address these challenges, we introduce FORESIGHT, a dual-stream architecture comprising two Siamese LLMs with shared weights, input encoders, and KV cache. The first LLM continuously processes incoming tokens, while the second runs ahead of the stream to anticipate future context, plan future computation, and generate task responses without interrupting streaming inference. Each plan decides when to reason next, what to check then, and how densely to sample, keeping transient evidence separate from persistent control. The resulting computation plan is executed online through an efficient reconfiguration protocol with schemaguided decoding and lightweight diff-based updates, enabling dynamic adaptation with low overhead. With a frozen Qwen3-VL-8B backbone, FORESIGHT achieves 23.0 mean joint F1 on OmniPro Online evaluation beating strongest trained baseline by 9.5%, while improving the backbone by 6.7 on StreamingBench and 15.4 on OVO-Bench, with the largest gain of 18.7 when evidence arrives later in the video stream. Our source code will be made publicly available.

Abstract: Video Object Segmentation (VOS) in complex and long videos is increasingly important for real-world applications, where target objects often appear only intermittently within long temporal horizons. However, existing benchmarks largely focus on temporally salient objects that remain visible for most of the video. To address this gap, we introduce FaVOS (A Benchmark for Video Object Segmentation with Fractional Temporal Visibility), a benchmark designed to evaluate VOS methods under low temporal visibility. We show that, in this regime, the standard J&F metric can collapse VOS evaluation into absence classification, because empty predictions receive high rewards on target-absent frames. Consequently, even a trivial empty-mask predictor can outperform strong models such as SAM 3, revealing a fundamental mismatch between current metrics and practical VOS performance. To mitigate this issue, we propose Volumetric J&F, which evaluates mask sequences as spatio-temporal volumes and reduces the dominance of target-absence rewards while preserving sensitivity to segmentation quality and temporal structure. Project page: https://aidaslab.github.io/FaVOS.

Abstract: Spatial reasoning benchmarks evaluate vision-language models across diverse tasks, but task-level scores do not reveal which underlying capabilities account for success or failure. Each task requires recovering spatial evidence, representing geometry, and reasoning over it. We disentangle these capabilities by comparing predicted and ground-truth spatial context under a shared schema and coordinate contract. This comparison reveals four recurring sources of error: inaccurate perception, missing information in the spatial context, selection of the wrong measurement, and errors in reference frames or in tracking position and orientation. Guided by this diagnosis, we develop CROSS, a training-free library of typed geometric operators and spatial skills that function over available evidence to support reliable video spatial reasoning. The resulting library supplies verified context to non-coding VLMs or callable skills to a SpatialClaw agent. We evaluate \methodname{} on five benchmarks. \methodname{} raises the average score from 55.9\% to 60.2\% on ReVSI and improves the SpatialClaw result from 62.8\% to 66.3\% on DSI-Bench. These gains demonstrate that explicit handling of spatial conventions can repair systematic reasoning failures without additional training.

Abstract: Video Anomaly Detection (VAD) aims to temporally localize abnormal events in videos. Most existing approaches rely on dataset-specific training and curated annotations, limiting generalization in open-set scenarios. Recent zero-shot methods based on Large Vision- Language Models (LVLMs) alleviate this dependency but often lack temporal continuity and structured reasoning. We propose Cog-VADU, a fully training-free framework that reformulates VAD as a sequential cognitive reasoning task. Cog-VADU introduces Chain-of- Anomaly Detection Thought Prompting (CoADTP), which unrolls an LVLM into a recurrent reasoning chain across video segments. By propagating structured rationales over time, the model maintains implicit temporal memory, enabling robust discrimination between com- plex anomalies and high-motion normal activities. To improve reliability, we further design a cross-modal re-ranking stage that aligns textual rationales with visual embeddings, enforcing semantic consistency and temporal coherence for refined and stable predictions. Extensive experiments on multiple public VAD benchmarks demonstrate that Cog-VADU achieves competitive zero-shot performance. Moreover, cross-model evaluations show that CoADTP consistently enhances reasoning-based anomaly detection in a model-agnostic manner, pro- viding interpretable and generalizable anomaly understanding for real-world applications.

Abstract: Long-video question answering is limited by the high cost of visual tokens and by the fixed context width of current VLMs. A long-video question may require broad temporal coverage, but the answer is often supported by only a compact set of moments. To locate these moments efficiently, we propose token-budgeted Video Evidence Indexing (VEI): given a dense low-resolution Video Preview, the model constructs a compact high-resolution Evidence Set for final reasoning. We treat VEI as a policy that must jointly solve \textit{evidence localization}, which finds question-relevant moments, and \textit{budget planning}, which decides where to spend the limited high-resolution frame budget. We implement this idea with an inference pipeline: the Video Preview provides cheap global coverage, Video Evidence Indexing constructs the Evidence Set, and Answer Generation combines both inputs for final VQA. To address missing frame-level supervision, we adopt privileged self-distillation, where an answer-aware teacher guides the normal test-time policy on student-generated indexing traces. We explore previews at 1, 6, 12, and 24 visual tokens per frame, training a single policy that supports all four resolutions. Experiments show that Video Evidence Indexing improves accuracy under limited visual budgets, and self-distillation further improves both QA accuracy and temporal evidence localization.

Abstract: Video super-resolution (VSR) is normally judged by PSNR and SSIM on clips that were downsampled bicubically, although in surveillance its purpose is to make faces and licence plates \emph{recognisable}. We present FANVIDv2, a benchmark that scores VSR by what a recognition pipeline can do with its output. FANVIDv2 provides $320\times180$ low-resolution (LR) clips with high-resolution (HR) references for 48 public figures (with one HR gallery image each) and 375 licence-plate clips covering 360 distinct plate strings. LR clips are generated with a randomised compound degradation (blur, resize jitter, sensor noise, JPEG compression, final downsampling) rather than bicubic downsampling alone. Two metrics score recognition \emph{inside} detections: FaceRecBox rewards a face only if it is localised and correctly identified, and TextRecBox scores plate transcriptions by normalised edit distance weighted by localisation quality. With a 2.3\,M-parameter VSR baseline (RCDM), FaceRecBox rises from 0.6864 to 0.7222, identity accuracy on matched faces from 84.35\% to 86.93\%, and TextRecBox from 0.3088 to 0.3667; a residual-map gated variant (RCDM-RMGF) reaches 0.3801 on plates. We describe the degradation model, the baseline architectures and the scorers in detail, and release annotations, metadata, download and degradation scripts and evaluation code.

Abstract: Reinforcement learning (RL) has substantially improved the reasoning ability of multimodal language models through verifiable rewards and increasingly fine-grainedvisual or temporal credit assignment. In video reasoning, however, current RL methods typically train with a fixed sparse frame budget: increasing the number of frames makes autoregressive rollouts expensive, while too few frames may miss temporally localized events and fine-grained visual details. We present \textbf{Frame Differential On-Policy Self-Distillation (FD-OPSD)}, which transfers the useful evidence of dense frame observations to a sparse frame policy during RL training. FD-OPSD compares the policy's token level preferences for the same sampled response under sparse and dense views, and distills the resulting frame differential signal without an external teacher or dense autoregressive rollout. The method preserves sparse-frame rollouts and leaves inference unchanged. Across Qwen2.5-VL-7B and Qwen3-VL-4B on six video reasoning benchmarks, FD-OPSD yields higher overall average performance than the strongest corresponding GRPO, T-GRPO, or Video-KTR baselines across the 16, 32, and 64 frame evaluation settings. These results show that dense visual evidence can be transferred selectively during training through token level self-distillation while retaining sparse frame rollouts and unchanged inference.

Abstract: Joint-embedding self-supervised learning typically combines an invariance objective across augmented views with additional mechanisms to prevent representational collapse. These objectives are often applied after a projection head, while downstream tasks use the backbone representation before the projector. We find that this mismatch does not necessarily prevent dimensional collapse in the backbone, which can retain low effective rank and potentially limit downstream transfer. To address this, we introduce SACReg, a spectral anti-collapse regularizer motivated by an analysis of $λ$-balance, which captures the relative scale of weight matrices across layers. In a two-layer linear network, we show that (i) $λ$-balance prevents collapse, and (ii) our regularizer applied to the backbone induces $λ$-balance. In the nonlinear case, this regularizer leads to anti-collapse as well and, in realistic architectures on ImageNet100, it empirically increases the representations' ranks. We apply SACReg to JEPA and propose $λ$-JEPA, which improves over LeJEPA and VISReg on ImageNet-1k classification and in average linear-probe transfer performance across eight downstream image datasets. On video self-supervised learning, $λ$-JEPA improves over LeVJEPA and V-JEPA 2 on the Something-Something-v2 and Kinetics-400 benchmarks. Code is available at https://github.com/berkerdemirel/lambda-jepa.

Abstract: Vision-Language Models (VLMs) are far more reliable at recognizing what appears in a video than at reasoning about its spatial and temporal properties, such as metric distances, object dimensions, and consistent object identities across frames. We present Geometric Code, a perception-to-geometry pipeline that computes explicit spatial structure from video and supplies it to VLMs as context to augment reasoning. A perception layer segments and classifies objects and recovers depth, camera pose, and intrinsics from monocular RGB video. A deterministic geometric engine then back-projects, merges, and cleans these outputs into a spatial code, including per-object positions, dimensions, counts, inter-object distances, appearance order, and room geometry. The code is serialized into VLMs' prompts, either alongside the video or replacing it entirely. Specifically, there is no component trained or fine-tuned in our approach. On VSI-Bench, augmenting 2B and 4B open models with the spatial code improves average accuracy by +4.1 points over the frames-only baseline, with the largest gains on numeric estimation tasks such as absolute distance (+24.1 points). The results suggest that explicitly computed geometry, delivered through the language channel, recovers spatial competence that small VLMs cannot extract from pixels alone.

Abstract: Fully supervised repetitive action counting (RAC) has achieved strong performance, but requires dense temporal annotations that are costly and difficult to scale. We propose TReViS, a self-supervised video synthesis framework that enables training RAC models without any repetition labels. TReViS estimates the underlying temporal repetition structure of an unlabeled video via a Temporal Self-Similarity Matrix, infers its cycle statistics, and synthesizes new training sequences that preserve realistic repetition patterns while introducing controlled temporal variability. These synthesized videos are paired with pseudo-labels and used to train existing RAC architectures from scratch. Across multiple datasets and backbones, TReViS consistently outperforms prior self-supervised methods and achieves performance competitive with several supervised baselines, while remaining fully label-free, demonstrating the effectiveness of structure-aware video synthesis for label-free RAC. The source code is available at https://github.com/yfqi/TReViS.

Abstract: Video foundation models now reach human-level accuracy on physical-reasoning benchmarks, yet such tasks require predicting unobserved physical outcomes. Do these models perform human-like forward simulation, or do they exploit statistical regularities in visible scenes? Accuracy alone cannot distinguish these strategies. We introduce a distributional evaluation framework that treats model seeds and human raters as populations, enabling comparison of consensus, uncertainty, and strategy. On the Physion benchmark, we evaluate three ViT-L architectures (V-JEPA2, VideoMAEv2, DINOv2). V-JEPA2 narrows the accuracy gap to ~1 percentage point (73.2% vs. 74.2%), yet model-human disagreement reaches 26.4%, far exceeding human-human disagreement (4.8%), with substantially lower agreement (kappa ~ 0.48 vs. 0.91). The divergence follows forward-simulation demands: models outperform humans on geometric reasoning (linking, +11.8 pp) but underperform on gravitational dynamics (rolling, -11.8 pp) and causal chains (dominoes, -10.5 pp). Strategy fingerprinting confirms all three architectures share non-human strategies while none aligns with humans. Attribution analysis suggests that unobservable outcome features, rather than visible scene properties, predict this divergence, consistent with models relying more on scene-level statistical regularities than on explicit forward simulation, a systematic divergence that accuracy alone cannot reveal. Code is available at https://github.com/fanhong-li/model-human-divergence.

Abstract: Video agents have made substantial progress in long-video understanding. Yet effective video-agent systems require costly, time-consuming manual design and trial and error. Current self-improvement methods either refine low-impact prompts, recombine predefined micro-tools, or struggle with convergence in harness optimization. To bridge this gap, we target high-impact video-tool with VideoResearcher, a training-free multi-agent framework that autonomously designs, tests, and refines tools for video understanding, like a human researcher. VideoResearcher operates through dual Solving and Evolving loops: it analyzes tool-use trajectories to identify capability gaps, coordinates specialized agents to develop and validate executable tools, and reuses evolved tools to strengthen evidence acquisition in subsequent video reasoning. Through iterative tool refinement and validation, it progressively strengthens evidence acquisition without updating model parameters. VideoResearcher achieves state-of-the-art performance among self-improving agents and approaches the human-designed upper bound, demonstrating a training-free paradigm for long-video understanding that expands agent capabilities through autonomous tool development while reducing costly manual engineering.

Abstract: The MUMU track of the 8th Large-scale Video Object Segmentation (LSVOS) Challenge requires a single unified multimodal model to jointly solve image tagging (Task A), open-vocabulary object detection (Task B), and English captioning (Task C) under strict resource constraints (<=0.5B parameters and <=8 GB peak GPU memory). We present CapMap-MS-TTA, a training-free submission built on Microsoft Florence-2-base (~231M parameters), combining caption keyword mapping with multi-scale flip test-time augmentation. Task C uses the native <DETAILED_CAPTION> pathway with length/token sanitization. Task A maps the same detailed caption into the official quality/scene/event vocabularies via an expanded keyword lexicon with whole-word matching and a lightweight expand-hints stage. Task B runs Florence-2 open detection (<OD>) with multi-scale and horizontal-flip test-time augmentation (TTA), followed by label-aware non-maximum suppression (NMS). Without fine-tuning, the system improves our reproduced Florence-2 baseline from 15.16 to a best public score of 16.4815, and ranks 3rd on the final MUMU leaderboard.

Abstract: Multimodal Large Language Models have demonstrated impressive video understanding, yet their ability to reason over long-form narratives is often masked by visual-centric evaluations and inefficient context processing. Existing benchmarks over-rely on visual heuristics while marginalizing auditory cues, effectively reducing models to "silent observers" that bypass genuine cross-modal reasoning. Moreover, standard dense sampling creates an evidence-context trade-off: increasing frames to capture evidence inevitably leads to attention distraction and token explosion. To bridge these gaps, we present Video-HolmesV2, a novel benchmark designed for Deep Audio-Visual Coupling. Unlike previous works, it enforces an Evidence-Based Evaluation, requiring models to justify answers with precise spatio-temporal audio-visual evidence, thereby reducing confounding effects of guessing and hallucinated evidence. To support this, we introduce: (1) a Multi-Model Cross-Verification pipeline to ensure task rigor; (2) a Spatio-temporal Evidence-Aware Metric for fine-grained calibration. Furthermore, we propose an Audio-Text Guided Token Compression framework. By fusing task intent with auditory anchors, our method distills high-value reasoning cues to mitigate long-context noise. In our evaluation, even strong proprietary models achieve below 60% accuracy, while our approach outperforms comparable open-source omni-models.

Abstract: Streaming video understanding requires models to process continuous multimodal input while maintaining temporal context. Existing evaluations are predominantly reactive: they query a model at a selected timestamp and therefore do not assess when it should respond. Proactive interaction instead requires monitoring a standing request, responding within an appropriate interval after the target event, and otherwise remaining silent. We introduce ProactiveBench, which evaluates models at one-second stream intervals without an explicit response cue. Its six subtasks vary trigger ambiguity and timing tolerance. Event Sensitivity geometrically combines response and silence rates on the same recording; four window-based subtasks distinguish early, in-window, and missed responses; and Duplicate Counting penalizes omissions and repetitions. Premature responses outnumber missed responses for four of the six evaluated systems, revealing a substantial gap in the temporal decision-making required for human-like interaction.

Abstract: Micro-actions are subtle, low-intensity non-verbal behaviors that provide cues to fine-grained human states, including emotions and intentions. Recognizing them remains difficult because they are brief, contain weak visual changes, and often exhibit similar motion patterns across categories. This paper addresses these challenges with a fine-grained micro-action recognition method that combines full fine-tuning of InternVideo2.5, hierarchical soft fusion, and a lightweight candidate-label reranker. For the long-tailed label distribution in MA-52, we use class-balanced sampling and inverse-frequency reweighting to reduce the effect of frequent classes during training. We fine-tune InternVideo2.5 end to end and attach coarse and group-conditional fine-grained classification heads to the shared video representation, improving the consistency between coarse and fine predictions. For ambiguous samples, the candidate-label reranker uses hard samples and video-label matching to focus on easily confused fine-grained actions. Experiments validate the proposed method, which achieves a 79.99% F1-mean on MA-52 and ranks first in the 3rd Micro-Action Analysis Grand Challenge at ACM Multimedia 2026.

Abstract: Timestamp-supervised action segmentation aims to segment and classify actions in untrimmed videos with a random frame annotated per action. Precisely localizing action boundaries from timestamp annotations is crucial for this setting, as it enables generating framewise pseudo-labels and applying the well-explored fully-supervised training. However, prevailing methods struggle with intrinsic uncertainty in boundary localization due to less discriminative features in action-transiting regions. This imprecise boundary estimation significantly reduces the stability and reliability of the generated pseudo-labels in ambiguous action-transiting regions, consequently resulting in performance deterioration of the trained segmentation models. In our paper, we introduce the boundary voting network that mitigates feature ambiguity by hierarchically propagating video-level global prior knowledge into local action-transiting regions. By generating key action representations as votes throughout the video and targeting action-transiting regions, all votes collaboratively contribute to action-transiting feature enhancement and boundary localization refinement. Extensive experiments demonstrate the effectiveness of our method on GTEA, 50Salads, and Breakfast datasets.

Abstract: Progress in video anomaly understanding (VAU) has long been limited by inherent deficiencies of real-world anomaly videos, which are hard to collect and offer little control over their content. Synthetic anomaly approaches partially alleviate data scarcity, yet their generation remains largely controlled at the category or prompt level. They also lack component-level verification of video-text consistency and provide insufficient hard normal samples near the normal-anomaly boundary. To address this, we present AnomalyCraft-700K, a component-controllable synthetic anomaly dataset for fine-grained VAU, containing over 40K videos and over 700K task-level textual annotations. From fine-grained semantic components and a progressive three-stage pipeline, we craft anomaly events that are richly detailed, semantically controlled, and temporally structured, and additionally construct per-category hard normal samples to prompt the model to discriminate based on anomaly semantics rather than surface visual cues. Moreover, using the components as verification units, AnomalyCraft-700K further performs component-wise correction of video-text discrepancies introduced during generation, providing reliable annotations with verified cross-modal alignment for six tasks that progress from anomaly detection, through anomaly retrieval and captioning, to fine-grained anomaly reasoning. Evaluations of widely used methods under both traditional and MLLM-based protocols demonstrate that AnomalyCraft-700K serves as an effective source of supervision, from anomaly detection to fine-grained anomaly understanding.