Videos

Recent advances in Vision-Language Models (VLMs) have substantially enhanced their ability across multimodal video understanding benchmarks spanning temporal, action, object, and spatial understanding. However, we identify a critical yet overlooked issue: when embedded on-screen text contradicts the visual scene, existing VLMs systematically hallucinate, prioritizing overlay textual semantics over the actual visual content. We define this phenomenon as Text Overlay-Induced Hallucination (TOIH). In this work, we propose VisualTextTrap, the first comprehensive benchmark, including large-scale human-validated samples with specifically designed evaluation metrics. In particular, we construct VisualTextTrap from widely-used public datasets using a scalable hybrid pipeline of VLMs assisted text generation and rigorous manual verification. The benchmark features 6,057 samples annotated across 88 fine-grained attributes within four dimensions, with hallucination intensity quantified on a five-level scale (L1--L5) that reflects the semantic contradiction between overlay text and visual reality. Moreover, we propose Visual Text Hallucination Mitigation Mixture-of-Experts (VTHM-MoE), a novel Vision-Text Disentanglement framework that employs a dual-encoder architecture. Concretely, four dimension-specialized expert modules spanning Temporal, Action, Object, and Spatial reasoning are first pre-trained to identify and leverage cross-modal discrepancies between textual semantics and actual video content. We develop an Adaptive Token Routing Strategy to enable dynamic expert allocation, conferring robust resistance to TOIH while preserving performance on uncontaminated videos. Extensive experiments conducted on our VisualTextTrap benchmark verify the effectiveness of VTHM-MoE, outperforming state-of-the-art counterparts with diverse video question answering tasks.

We propose PPEDCRF, a calibrated selective perturbation framework that protects \emph{background-based location privacy} in released video frames against gallery-based retrieval attackers. Even after GPS metadata are stripped, an adversary can geolocate a frame by matching its background visual cues to geo-tagged reference imagery; PPEDCRF mitigates this threat by estimating location-sensitive background regions with a dynamic conditional random field (DCRF), rescaling perturbation strength with a normalized control penalty (NCP), and injecting Gaussian noise only inside the inferred regions via a DP-style calibration rule. On a controlled paired-scene retrieval benchmark with eight attacker backbones and three noise seeds, PPEDCRF reduces ResNet18 Top-1 retrieval accuracy from 0.667 to $0.361\pm0.127$ at $σ_0=8$ while preserving $36.14\,$dB PSNR -- an ${\approx}6\,$dB quality advantage over global Gaussian noise. Transfer across the eight-backbone seed-averaged benchmark is broadly supportive (23 of 24 backbone-gallery cells show negative $Δ$), while appendix-scale confirmation identifies MixVPR as a remaining adverse-transfer exception. Matched-operating-point analysis shows that PPEDCRF and global Gaussian noise converge in Top-1 privacy at equal utility, so the practical benefit is spatially concentrated perturbation that preserves higher visual quality at any given noise scale rather than stronger matched-utility privacy. Code: https://github.com/mabo1215/PPEDCRF

Zero-shot action recognition is challenging due to the semantic gap between seen and unseen classes. We present a novel framework that enhances CLIP with disentangled embeddings and semantic-guided interaction. A Motion Separation Module (MSM) separates motion-sensitive and global-static features, while a Motion Aggregation Block (MAB) employs gated cross-attention to refine motion representation without re-coupling redundant information. To facilitate generalization to unseen categories, we enforce semantic alignment between video features and textual representations by aligning projected embeddings with positive textual prompts, while leveraging negative prompts to explicitly model "non-class" semantics. Experiments on standard benchmarks demonstrate that our method consistently outperforms prior CLIP-based approaches, achieving robust zero-shot action recognition across both coarse and fine-grained datasets.

Streaming video reasoning requires models to operate in a setting where history grows without bound while meaningful evidence remains scarce. In such a landscape, relevant signal is like an oasis-small, critical, and easily lost in a desert of redundancy. Enlarging memory only widens the desert; aggressive compression dries up the oasis. The real difficulty lies in discovering where to look, not how much to remember. We therefore introduce OASIS, a novel framework for streaming video reasoning that tackles this challenge through structured, on-demand retrieval. It organizes streaming history into hierarchical events and performs reasoning as controlled refinement-short-context inference first, followed by semantically grounded retrieval only when uncertainty arises. As the retrieval is driven by high-level intent rather than embedding similarity, the retrieved memory is substantially more accurate and less noisy. Additionally, the mechanism is plug-and-play, training-free, and readily attaches to different streaming MLLM backbones. Experiments across multiple benchmarks and backbones show that OASIS achieves strong gains in long-horizon accuracy and compositional reasoning with bounded token cost and low request delay. Code is available at https://github.com/Solus-sano/OASIS.

Video editing aims to modify input videos according to user intent. Recently, end-to-end training methods have garnered widespread attention, constructing paired video editing data through video generation or editing models. However, compared to image editing, the high annotation costs of video data severely constrain the scale, quality, and task diversity of video editing datasets when relying on video generative models or manual annotation. To bridge this gap, we propose LIVE, a joint training framework that leverages large-scale, high-quality image editing data alongside video datasets to bolster editing capabilities. To mitigate the domain discrepancy between static images and dynamic videos, we introduce a frame-wise token noise strategy, which treats the latents of specific frames as reasoning tokens, leveraging large pretrained video generative models to create plausible temporal transformations. Moreover, through cleaning public datasets and constructing an automated data pipeline, we adopt a two-stage training strategy to anneal video editing capabilities. Furthermore, we curate a comprehensive evaluation benchmark encompassing over 60 challenging tasks that are prevalent in image editing but scarce in existing video datasets. Extensive comparative and ablation experiments demonstrate that our method achieves state-of-the-art performance. The source code will be publicly available.

Reinforcement learning from verifiable rewards (RLVR) has demonstrated remarkable effectiveness in improving the reasoning capabilities of large language models. As models evolve into natively multimodal architectures, extending RLVR to video understanding becomes increasingly important yet remains largely unexplored, due to the diversity of video task types, the computational overhead of repeatedly decoding and preprocessing high-dimensional visual inputs, and the difficulty of reproducible evaluation across numerous sensitive hyperparameters. Existing open-source RL training frameworks provide solid infrastructure for text and image scenarios but lack systematic optimizations tailored for video modality. In this work, we present \textbf{EasyVideoR1}, a complete and efficient reinforcement learning framework specifically designed for training large vision-language models on video understanding tasks. EasyVideoR1 makes the following contributions: (1) a full video RL training pipeline with offline preprocessing and tensor caching that eliminates redundant video decoding and yields a 1.47 $\times$ throughput improvement; (2) a comprehensive, task-aware reward system covering 11 distinct video and image problem types with unified routing and modular extension; (3) a mixed offline-online data training paradigm that combines curated high-quality trajectories with on-policy exploration, benefiting the learning of more challenging tasks; (4) joint image-video training with independently configurable pixel budgets, allowing the two modalities to mutually reinforce each other; and (5) an asynchronous multi-benchmark evaluation framework covering 22 mainstream video understanding benchmarks, with reproduced accuracy closely aligned with officially reported scores.

Action recognition is a fundamental ability for social species. Yet, its underlying computations are not well understood. Classical psychophysical studies using simplified stimuli have shown that humans can perceive body motion even under degradation of relevant shape cues. Recent work using real-world action videos and their appearance-free counterparts (that preserve motion but lack static shape cues) included explicit training of humans and models on the appearance-free videos. Whether humans and vision models generalize in a zero-shot manner to appearance-free transformations of real-world action videos is not yet known. To measure this generalization in humans, we conducted a laboratory-based psychophysics experiment. 22 participants were trained to recognize five action categories using naturalistic videos (UCF5 dataset), and tested zero-shot on two types of appearance-free transformations: (i) dense-noise motion videos from an existing dataset (AFD5) and (ii) random-dot appearance-free videos. We find that participants recognize actions in both types of appearance-free videos well above chance, albeit with reduced accuracy compared to naturalistic videos. To model this behavior, we developed a two-pathway 3D CNN-based model combining an RGB (form) stream and an optical flow (motion) stream, including a coherence-gating mechanism inspired by Gestalt common-fate grouping. Our model generalizes to both appearance-free datasets and outperforms contemporary video classification models, narrowing the gap to human performance. We find that the motion pathway is critical for generalization to appearance-free videos, while the form pathway improves performance on naturalistic videos. Our findings highlight the importance of motion-based representations for generalization to appearance-free videos, and support the use of multi-stream architectures to model video-based action recognition.

Time-to-Collision (TTC) forecasting is a critical task in collision prevention, requiring precise temporal prediction and comprehending both local and global patterns encapsulated in a video, both spatially and temporally. To address the multi-scale nature of video, we introduce a novel spatiotemporal hierarchical transformer-based architecture called CollideNet, specifically catered for effective TTC forecasting. In the spatial stream, CollideNet aggregates information for each video frame simultaneously at multiple resolutions. In the temporal stream, along with multi-scale feature encoding, CollideNet also disentangles the non-stationarity, trend, and seasonality components. Our method achieves state-of-the-art performance in comparison to prior works on three commonly used public datasets, setting a new state-of-the-art by a considerable margin. We conduct cross-dataset evaluations to analyze the generalization capabilities of our method, and visualize the effects of disentanglement of the trend and seasonality components of the video data. We release our code at https://github.com/DeSinister/CollideNet/.

Understanding affective dynamics in real-world social systems is fundamental to modeling and analyzing human-human interactions in complex environments. Group affect emerges from intertwined human-human interactions, contextual influences, and behavioral cues, making its quantitative modeling a challenging computational social systems problem. However, computational modeling of group affect in in-the-wild scenarios remains challenging due to limited large-scale annotated datasets and the inherent complexity of multimodal social interactions shaped by contextual and behavioral variability. The lack of comprehensive datasets annotated with multimodal and contextual information further limits advances in the field. To address this, we introduce the Group Affect from ViDeos (GAViD) dataset, comprising 5091 video clips with multimodal data (video, audio and context), annotated with ternary valence and discrete emotion labels and enriched with VideoGPT-generated contextual metadata and human-annotated action cues. We also present Context-Aware Group Affect Recognition Network (CAGNet) for multimodal context-aware group affect recognition. CAGNet achieves 63.20% test accuracy on GAViD, comparable to state-of-the-art performance. The dataset and code are available at github.com/deepakkumar-iitr/GAViD.

The evaluation of visual editing models remains fragmented across methods and modalities. Existing benchmarks are often tailored to specific paradigms, making fair cross-paradigm comparisons difficult, while video editing lacks reliable evaluation benchmarks. Furthermore, common automatic metrics often misalign with human preference, yet directly deploying large multimodal models (MLLMs) as evaluators incurs prohibitive computational and financial costs. We present UniEditBench, a unified benchmark for image and video editing that supports reconstruction-based and instruction-driven methods under a shared protocol. UniEditBench includes a structured taxonomy of nine image operations (Add, Remove, Replace, Change, Stroke-based, Extract, Adjust, Count, Reorder) and eight video operations, with coverage of challenging compositional tasks such as counting and spatial reordering. To enable scalable evaluation, we distill a high-capacity MLLM judge (Qwen3-VL-235B-A22B Instruct) into lightweight 4B/8B evaluators that provide multi-dimensional scoring over structural fidelity, text alignment, background consistency, naturalness, and temporal-spatial consistency (for videos). Experiments show that the distilled evaluators maintain strong agreement with human judgments and substantially reduce deployment cost relative to the teacher model. UniEditBench provides a practical and reproducible protocol for benchmarking modern visual editing methods. Our benchmark and the associated reward models are publicly available at https://github.com/wesar1/UniEditBench.

Reliable procedural monitoring in video requires exposure to naturally occurring human errors and the recoveries that follow. In egocentric recordings, mistakes are often partially occluded by hands and revealed through subtle object state changes, while existing procedural datasets provide limited and inconsistent mistake and correction traces. We present PIE-V (Psychologically Inspired Error injection for Videos), a framework for constructing and benchmarking mistake-aware egocentric procedural videos by augmenting clean keystep procedures with controlled, human-plausible deviations. PIE-V combines a psychology-informed error planner conditioned on procedure phase and semantic step load, a correction planner that models recovery behavior, an LLM writer that performs cascade-consistent rewrites, and an LLM judge that validates procedural coherence and repairs failures. For video segment edits, PIE-V synthesizes replacement clips with text-guided video generation and stitches them into the episode to preserve visual plausibility. Applied to 17 tasks and 50 Ego-Exo4D scenarios, PIE-V injects 102 mistakes and generates 27 recovery corrections. For benchmarking, we introduce a unified taxonomy and a human rubric with nine metrics that cover step-level and procedure-level quality, including plausibility, procedure logic with annotator confidence, state change coherence, and grounding between text and video. Using this protocol, we audit several existing resources and compare PIE-V against a freeform LLM generation baseline under the same criteria. Together, the framework and rubric support post-completion verification for egocentric procedural mistake detection and correction.

This paper presents an overview of the NTIRE 2026 Challenge on Video Saliency Prediction. The goal of the challenge participants was to develop automatic saliency map prediction methods for the provided video sequences. The novel dataset of 2,000 diverse videos with an open license was prepared for this challenge. The fixations and corresponding saliency maps were collected using crowdsourced mouse tracking and contain viewing data from over 5,000 assessors. Evaluation was performed on a subset of 800 test videos using generally accepted quality metrics. The challenge attracted over 20 teams making submissions, and 7 teams passed the final phase with code review. All data used in this challenge is made publicly available - https://github.com/msu-video-group/NTIRE26_Saliency_Prediction.

Video understanding requires identifying and reasoning over semantically discriminative visual objects across frames, yet existing object-agnostic solutions struggle to effectively handle substantial object variations over time. To address this, we introduce Chain-of-Glimpse, a search-guided progressive object-grounded reasoning framework that explicitly anchors each reasoning step to specific visual evidence regions, enabling compositional and multi-step decision-making. Formally, Chain-of-Glimpse formulates video reasoning as a step-by-step process that incrementally builds spatially grounded traces around task-relevant visual objects, thereby mitigating over-reliance on saliency-driven cues. Specifically, Chain-of-Glimpse features a search-guided controller, optimized via reinforcement learning with a format reward that significantly incentivizes grounding capability, to iteratively ground visual evidence regions and form reliable reasoning trajectories, yielding accurate and interpretable multi-step decisions. Extensive evaluations on both in domain NExTQA and out-of-domain Video-Holmes, CG-Bench Reasoning, and VRBench benchmarks demonstrate consistent performance gains, robustness and generalization of Chain-of-Glimpse across diverse video reasoning tasks.

Training multimodal large language models (MLLMs) for video understanding requires large-scale annotated data spanning diverse tasks such as object counting, question answering, and segmentation. However, collecting and annotating multimodal video data in real-world is costly, slow, and inherently limited in diversity and coverage. To address this challenge, we propose~\textbf{SynMulti} dataset, along with a unified synthetic data generation pipeline capable of automatically producing unlimited multimodal video data with rich and diverse supervision. Our framework supports multiple task formats within a single pipeline, enabling scalable and consistent data creation across tasks. To further enhance reasoning ability, we introduce a VQA-based fine-tuning strategy that trains models to answer structured questions about visual content rather than relying solely on captions or simple instructions. This formulation encourages deeper visual grounding and reasoning. We evaluate our approach in three challenging tasks: video object counting, video-based visual question answering, and video object segmentation. Experimental results demonstrate that models trained predominantly on synthetic data generalize effectively to real-world datasets, often outperforming traditionally trained counterparts. Our findings highlight the potential of unified synthetic data pipelines as a scalable alternative to expensive real-world annotation for multimodal video understanding.

Using behavioural science, health interventions focus on behaviour change by providing a framework to help patients acquire and maintain healthy habits that improve medical outcomes. In-person interventions are costly and difficult to scale, especially in resource-limited regions. Digital health interventions offer a cost-effective approach, potentially supporting independent living and self-management. Automating such interventions, especially through machine learning, has recently gained considerable attention. Ambivalence and hesitancy (A/H) play a primary role for individuals to delay, avoid, or abandon health interventions. A/H are subtle and conflicting emotions that place a person in a state between positive and negative evaluations of a behaviour, or between acceptance and refusal to engage in it. They manifest as affective inconsistency across modalities or within a modality, such as language, facial, vocal expressions, and body language. While experts can be trained to recognize A/H, integrating them into digital health interventions is costly and less effective. Automatic A/H recognition is therefore critical for the personalization and cost-effectiveness of digital health interventions. Here, we explore the application of deep learning models for A/H recognition in videos, a multi-modal task by nature. In particular, this paper covers three learning setups: supervised learning, unsupervised domain adaptation for personalization, and zero-shot inference via large language models (LLMs). Our experiments are conducted on the unique and recently published BAH video dataset for A/H recognition. Our results show limited performance, suggesting that more adapted multi-modal models are required for accurate A/H recognition. Better methods for modeling spatio-temporal and multimodal fusion are necessary to leverage conflicts within/across modalities.

Game engines hold what video models struggle to learn: a complete, explicit world state behind every frame. We turn one into a data instrument. GEST-Engine, our production-grade open-source system, deterministically executes Graphs of Events in Space and Time (GESTs), whether procedurally generated or derived from text, into videos of synchronized multi-actor scenarios, recording ground truth as it renders: 3D entity and camera state, pairwise spatial relations, event-to-frame mappings, instance segmentation, and long descriptions, at zero marginal annotation cost. With it we release GTASA, a 938-video sample of what the system can generate at arbitrary scale, carrying, to our knowledge, the densest spatial-relation coverage of any video dataset: a complete entity-pair relation graph at every frame, ~84x denser than the state of the art, frame-for-frame. We validate GTASA both qualitatively, through human evaluation of physical validity and semantic alignment where frontier neural generators, given the same prompts, largely fail, and quantitatively, with GTASA pretraining improving VLM video captioning. Probing six frozen video encoders across 11 spatio-temporal tasks enabled by GTASA's exact 3D ground truth, a previously untestable inter-entity relational probe of frozen video features, reveals that who-is-near-whom barely rises above chance for all of them. We release the engine, the corpus, and the benchmark, making this gap a measurable, trainable target.

Multimodal fake news video detection is a crucial research direction for maintaining the credibility of online information. Existing studies primarily verify content authenticity by constructing multimodal feature fusion representations or utilizing pre-trained language models to analyze video-text consistency. However, these methods still face the following limitations: (1) lacking cross-instance global semantic correlations, making it difficult to effectively utilize historical associative evidence to verify the current video; (2) semantic discrepancies across domains hinder the transfer of general knowledge, lacking the guidance of domain-specific expert knowledge. To this end, we propose a novel Retrieval-Augmented Semantic Reasoning (RASR) framework. First, a Cross-instance Semantic Parser and Retriever (CSPR) deconstructs the video into high-level semantic primitives and retrieves relevant associative evidence from a dynamic memory bank. Subsequently, a Domain-Guided Multimodal Reasoning (DGMP) module incorporates domain priors to drive an expert multimodal large language model in generating domain-aware, in-depth analysis reports. Finally, a Multi-View Feature Decoupling and Fusion (MVDFF) module integrates multi-dimensional features through an adaptive gating mechanism to achieve robust authenticity determination. Extensive experiments on the FakeSV and FakeTT datasets demonstrate that RASR significantly outperforms state-of-the-art baselines, achieves superior cross-domain generalization, and improves the overall detection accuracy by up to 0.93%.

Continuous inference over concurrent video streams imposes substantial compute and memory demands on vision-language model (VLM) serving. Streaming inference uses sliding windows to maintain a bounded context of recent video, but processing each window independently repeats visual encoding and large language model (LLM) prefilling for similar and overlapping content. Existing optimizations provide limited coordination across these stages and often rely on model-specific training, profiling, or model-generated signals. We present CodecSight, a streaming VLM serving system that uses codec metadata as shared runtime guidance across visual encoding and LLM prefilling, without model-specific training or offline profiling. Codec-derived change signals guide patch pruning before visual encoding, reducing both visual computation and the number of downstream visual tokens. Codec-defined frame types guide selective key-value (KV) refresh across windows, while positional correction enables reuse of the remaining cached keys. Across three VLMs and four video workloads, our vLLM-based implementation supports up to $3.3\times$ as many concurrent streams and achieves up to a $5.3\times$ speedup in average time-to-first-token relative to the state-of-the-art baselines. It also reduces executed FLOPs by up to 93%, with a maximum task-quality decrease of 4.64 percentage points.

Automatic violence detection from video is challenging because violent interactions may be distant, occluded, or only partially visible. Audio can provide complementary evidence for violent events that are difficult to recognize from visual information alone. However, audio itself may be absent, dubbed, or dominated by environmental noise, making the central challenge not whether to incorporate audio but how to adapt reliance on it according to the visual scene. We introduce \emph{AViS-Mamba}, an audiovisual Mamba-based architecture in which the visual stream directly governs the behavior of the audio stream. At each layer of the audio encoder, a compact visual representation produces a modulation vector that conditions the encoder's internal temporal operators together with a routing gate that regulates the strength of this visual intervention. Rather than fusing or reweighting features after they have been extracted, visual context directly shapes the temporal dynamics of the audio encoder. We further propose Adaptive AV-InfoNCE, a contrastive objective that learns to balance the audio-to-video and video-to-audio alignment directions rather than weighting them uniformly. On the audio-valid NTU-CCTV and DVD benchmarks, AViS-Mamba establishes state-of-the-art results, attaining 88.59% and 75.74% accuracy. We demonstrate that adaptive visual conditioning consistently outperforms fixed routing and improves performance under degraded and missing-audio conditions. Layer-wise analysis further reveals that the model adapts the audio stream selectively across network depth rather than applying a single global routing policy.

Surgical video understanding is a crucial prerequisite for advancing Computer-Assisted Surgery. While vision-language models (VLMs) have recently been applied to the surgical domain, existing surgical vision-language datasets lack in capturing and evaluating complex, interleaved spatial-temporal dynamics. Creating large scale datasets that accurately represent fine-grained spatial-temporal relationships in surgical videos is challenging due to costly manual annotations or error-prone generation using large language models. To address this gap, we introduce the SurgSTU-Pipeline, a deterministic generation pipeline featuring temporal and spatial continuity filtering to reliably create surgical datasets for fine-grained spatial-temporal multimodal understanding. Applying this pipeline to publicly available surgical datasets, we create the SurgSTU dataset, comprising 6711 video clips densely extended with 150k fine-grained spatial-temporal question-answer samples. Our comprehensive evaluation shows that while state-of-the-art generalist VLMs struggle in zero-shot settings, their spatial-temporal capabilities can be improved through in-context learning. A fine-tuned VLM on the SurgSTU training dataset achieves highest performance among all spatial-temporal tasks, validating the dataset's efficacy to improve spatial-temporal understanding of VLMs in surgical videos. The project is available here: https://lennart-maack.github.io/SurgSTU-project

The inherent complexity of video understanding makes it difficult to determine whether Video-LLM benchmark performance stems from visual perception, linguistic reasoning, or knowledge priors. While many benchmarks have emerged to assess high-level reasoning, shared criteria for evaluating video understanding remain largely overlooked. Instead of introducing yet another benchmark, we take a step back to re-examine the criteria for evaluating video understanding. In this work, we introduce Video-Oasis, a sustainable diagnostic suite for systematically auditing existing video understanding benchmarks. This audit reveals that 55% of existing benchmark samples are solvable without visual input or temporal context. After filtering these shortcuts, the remaining video-native challenges expose a substantial capability gap: state-of-the-art models perform only marginally above random guessing. Building on these findings, we use the distilled challenges as a testbed to investigate which algorithmic design choices contribute to robust video understanding. We hope our work provides a practical foundation for constructing rigorous video benchmarks and evaluating future Video-LLMs. Code is available at https://github.com/sejong-rcv/Video-Oasis.

Generative video editing has enabled creative control over an object's position in a video by prescribing an object's 3D or 2D motion trajectory, while preserving both video plausibility and identity. However, manually specifying a plausible full motion trajectory, like the arcs of a bouncing ball, requires time and expertise, and may therefore not be a suitable editing task for non-experts or quick edits. In contrast, in the image domain, generative object translation has been established as a simple editing task that requires only a single drag to move an object to a new location; yet an equivalent method is still missing for videos. We propose an analogous new editing task for videos that translates an object's 3D motion trajectory in a video while preserving identity and plausibility of appearance and motion, for example, translating the trajectory of a bouncing ball while preserving its relative motion. The main challenge in training this task lies in obtaining paired video data for this scenario. We introduce TrajectorySynth, a new data generation strategy for large-scale synthetic paired video data and a video generator TrajectoryMover fine-tuned with this data. We show that this enables generative movement of object trajectories. Project Page: https://chhatrekiran.github.io/trajectorymover

Early screening via colonoscopy is critical for colon cancer prevention, yet developing robust AI systems for this domain is hindered by the lack of densely annotated, long-sequence video datasets. Existing datasets predominantly focus on single-class polyp detection and lack the rich spatial, temporal, and linguistic annotations required to evaluate modern Multimodal Large Language Models (MLLMs). To address this critical gap, we introduce Colon-Bench, generated via a novel multi-stage agentic workflow. Our pipeline seamlessly integrates temporal proposals, bounding-box tracking, AI-driven visual confirmation, and human-in-the-loop review to scalably annotate full-procedure videos. The resulting verified benchmark is unprecedented in scope, encompassing 528 videos, 14 distinct lesion categories (including polyps, ulcers, and bleeding), over 300,000 bounding boxes, 213,000 segmentation masks, and 133,000 words of clinical descriptions. We utilize Colon-Bench to rigorously evaluate state-of-the-art MLLMs across lesion classification, Open-Vocabulary Video Object Segmentation (OV-VOS), and video Visual Question Answering (VQA). The MLLM results demonstrate surprisingly high localization performance in medical domains compared to SAM-3. Finally, we analyze common VQA errors from MLLMs to introduce a novel "colon-skill" prompting strategy, improving zero-shot MLLM performance by up to 9.7% across most MLLMs. The dataset and the code are available at https://abdullahamdi.com/colon-bench .

Existing Multimodal Large Language Models (MLLMs) often suffer from hallucinations in long video understanding (LVU), primarily due to the imbalance between textual and visual tokens. Observing that MLLMs handle short visual inputs well, recent LVU works alleviate hallucinations by automatically parsing the vast visual data into manageable segments that can be effectively processed by MLLMs. SFT-based tool-calling methods can serve this purpose, but they typically require vast amounts of fine-grained, high-quality data and suffer from constrained tool-calling trajectories. We propose a novel VideoTIR that leverages Reinforcement Learning (RL) to encourage proper usage of comprehensive multi-level toolkits for efficient long video understanding. VideoTIR explores both Zero-RL and SFT cold-starting to enable MLLMs to retrieve and focus on meaningful video segments/images/regions, enhancing long video understanding both accurately and efficiently. To reduce redundant tool-calling, we propose Toolkit Action Grouped Policy Optimization (TAGPO), which enhances the efficiency of the calling process through stepwise reward assignment and reuse of failed rollouts. Additionally, we develop a sandbox-based trajectory synthesis framework to generate high-quality trajectories data. Extensive experiments on three long-video QA benchmarks demonstrate the effectiveness and efficiency of our method.

Automated semantic annotation of broadcast television content presents distinctive challenges, combining structured audiovisual composition, domain-specific editorial patterns, and strict operational constraints. While multimodal large language models (MLLMs) have demonstrated strong general-purpose video understanding capabilities, their comparative effectiveness across pipeline architectures and input configurations in broadcast-specific settings remains empirically undercharacterized. This paper presents a systematic evaluation of multimodal annotation pipelines applied to broadcast television news in the Italian setting. We construct a domain-specific benchmark of clips labeled across four semantic dimensions: visual environment classification, topic classification, sensitive content detection, and named entity recognition. Two different pipeline architectures are evaluated across nine frontier models, including Gemini 3.0 Pro, LLaMA 4 Maverick, Qwen-VL variants, and Gemma 3, under progressively enriched input strategies combining visual signals, automatic speech recognition, speaker diarization, and metadata. Experimental results demonstrate that gains from video input are strongly model-dependent: larger models effectively leverage temporal continuity, while smaller models show performance degradation under extended multimodal context, likely due to token overload. Beyond benchmarking, the selected pipeline is deployed on 14 full broadcast episodes, with minute-level annotations integrated with normalized audience measurement data provided by an Italian media company. This integration enables correlational analysis of topic-level audience sensitivity and generational engagement divergence, demonstrating the operational viability of the proposed framework for content-based audience analytics.

Despite decades of work, surveillance still struggles in searching and reasoning about specific targets across long, multi-camera videos. Existing methods - tracking, retrieval, and video LLMs require heavy manual filtering, capture only shallow attributes, and fail at temporal understanding. Prior benchmarks are also limited to basic retrieval and question answering, without addressing real world challenges that often involve multimodal queries and temporal grounding (e.g., "When did this person join the fight?" with the person's image). To address this gap, we introduce ForeSeaQA, a new benchmark specifically designed for video QA with image-and-text queries and timestamped annotations of key events. The dataset consists of long-horizon surveillance footage paired with diverse multimodal questions, enabling systematic evaluation of retrieval, temporal grounding, and multimodal reasoning in realistic forensic conditions. Not limited to this benchmark, we propose ForeSea, an AI forensic search system with a 3-stage, plug-and-play pipeline. (1) A tracking module filters irrelevant footage; (2) a multimodal embedding module indexes the remaining clips; and (3) during inference, the system retrieves top-K candidate clips for a video LLM to answer queries and localize events. On ForeSeaQA benchmark, ForeSea improves accuracy by 3.1 points and temporal IoU by 10.1 points over prior retrieval-augmented baselines. To our knowledge, ForeSeaQA is the first benchmark to support complex multimodal queries with precise temporal grounding, and ForeSea is the first VideoRAG system built to excel in this setting.

The ubiquity of multimedia content is reshaping online information spaces, particularly in social media environments. At the same time, search is being rapidly transformed by generative AI, with large language models (LLMs) routinely deployed as intermediaries between users and multimedia content to retrieve and summarize information. Despite their growing influence, the impact of LLM inaccuracies and potential vulnerabilities on multimedia information-seeking tasks remains largely unexplored. We investigate how generative AI affects accuracy, efficiency, and confidence in information retrieval from videos. We conduct an experiment with around 900 participants on 8,000+ video-based information-seeking tasks, comparing behavior across three conditions: (1) access to videos only, (2) access to videos with LLM-based AI assistance, and (3) access to videos with a deceiving AI assistant designed to provide false answers. We find that AI assistance increases accuracy by 3-7% when participants viewed the relevant video segment, and by 27-35% when they did not. Efficiency increases by 10% for short videos and 25% for longer ones. However, participants tend to over-rely on AI outputs, resulting in accuracy drops of up to 32% when interacting with the deceiving AI. Alarmingly, self-reported confidence in answers remains stable across all three conditions. Our findings expose fundamental safety risks in AI-mediated video information retrieval.

Recent advances in Streaming Video Understanding has enabled a new interaction paradigm where models respond proactively to user queries. Current proactive VideoLLMs rely on per-frame triggering decision making, which suffers from an efficiency-accuracy dilemma. We propose Em-Garde, a novel framework that decouples semantic understanding from streaming perception. At query time, the Instruction-Guided Proposal Parser transforms user queries into structured, perceptually grounded visual proposals; during streaming, a Lightweight Proposal Matching Module performs efficient embedding-based matching to trigger responses. Experiments on StreamingBench and OVO-Bench demonstrate consistent improvements over prior models in proactive response accuracy and efficiency, validating an effective solution for proactive video understanding under strict computational constraints.

Long-form video question answering requires reasoning over extended temporal contexts, making frame selection a critical bottleneck for multi-modal large language models (MLLMs) bound by finite context windows. Within the controlled frame-budget regime that governs practical deployment, prior selectors score frames against a single global query embedding; as a result, compositional multimodal questions that involve temporal ordering or cross-modal cues such as ``what happens on screen right after the narrator mentions the reaction?'' are flattened into a representation that loses sub-event ordering and modality bindings. We introduce \textbf{HiMu}, a training-free framework for compositional multimodal frame selection. A single text-only LLM call decomposes the query into a hierarchical logic tree whose leaves are atomic predicates, each routed to a lightweight expert spanning vision (CLIP, open-vocabulary detection, OCR) and audio (speech recognition and non-speech sound matching). Expert signals are normalized, smoothed to align across modalities, and composed bottom-up through fuzzy-logic operators that enforce temporal sequencing and adjacency, yielding a continuous per-frame satisfaction curve. Under the standard 16-frame budget on Video-MME, LongVideoBench, and HERBench-Lite, HiMu achieves state-of-the-art accuracy among frame selection methods and improves over uniform sampling across seven diverse MLLMs as a drop-in module, matching the accuracy of uniform sampling at $4\times$ its frame budget, without retraining and without multiple iterative MLLM calls during selection.

We present AHOY, a method for reconstructing complete, animatable 3D Gaussian avatars from in-the-wild monocular video despite heavy occlusion. Existing methods assume unoccluded input-a fully visible subject, often in a canonical pose-excluding the vast majority of real-world footage where people are routinely occluded by furniture, objects, or other people. Reconstructing from such footage poses fundamental challenges: large body regions may never be observed, and multi-view supervision per pose is unavailable. We address these challenges with four contributions: (i) a hallucination-as-supervision pipeline that uses identity-finetuned diffusion models to generate dense supervision for previously unobserved body regions; (ii) a two-stage canonical-to-pose-dependent architecture that bootstraps from sparse observations to full pose-dependent Gaussian maps; (iii) a map-pose/LBS-pose decoupling that absorbs multi-view inconsistencies from the generated data; (iv) a head/body split supervision strategy that preserves facial identity. We evaluate on YouTube videos and on multi-view capture data with significant occlusion and demonstrate state-of-the-art reconstruction quality. We also demonstrate that the resulting avatars are robust enough to be animated with novel poses and composited into 3DGS scenes captured using cell-phone video. Our project page is available at https://miraymen.github.io/ahoy/

Video understanding requires models to continuously track and update world state during playback. Although existing benchmarks have advanced video understanding evaluation across multiple dimensions, they provide limited visibility into how models maintain world state over time. We propose SVCBench, a Streaming Video Counting Benchmark that repositions counting as a minimal, controlled probe for diagnosing models' world-state maintenance capability. We decompose this capability into object counting and event counting, forming 8 fine-grained subcategories. Object counting covers tracking currently visible objects and cumulative unique identities, while event counting covers detecting instantaneous actions and tracking complete activity cycles. SVCBench contains 406 videos with frame-by-frame annotations of 10,071 event occurrences and object state changes, yielding 1,000 streaming QA pairs with 4,576 query points distributed along video timelines. By observing state maintenance trajectories through streaming multi-point queries, we design three complementary metrics to diagnose numerical precision, trajectory consistency, and temporal awareness. Evaluations of mainstream video-language models show that current models still exhibit significant deficiencies in spatial-temporal state maintenance, with especially poor performance on periodic event counting. SVCBench provides a diagnostic framework for measuring and improving state maintenance in video understanding systems. Our code and data are available at https://buaa-colalab.github.io/SVCBench.

Multimodal instruction tuning is often compute-inefficient because training budgets are spread across large mixed image-video pools whose utility is highly uneven. We present Goal-Driven Data Optimization (GDO), a framework that computes six sample descriptors for each candidate and constructs optimized 1$\times$ training subsets for different goals. Under a fixed one-epoch Qwen3-VL-8B-Instruct training and evaluation recipe on 8 H20 GPUs, GDO uses far fewer training samples than the Uni-10x baseline while converging faster and achieving higher accuracy. Relative to the fixed 512k-sample Uni-10x baseline, GDO reaches the Uni-10x reference after 35.4k samples on MVBench, 26.6k on VideoMME, 27.3k on MLVU, and 34.7k on LVBench, while improving Accuracy by +1.38, +1.67, +3.08, and +0.84 percentage points, respectively. The gains are largest on MVBench and MLVU, while LVBench improves more modestly, consistent with its ultra-long-video setting and the mismatch between that benchmark and the short-video/image-dominant training pool. Across MinLoss, Diverse, Temp, and Temp+, stronger temporal emphasis yields steadily better long-video understanding behavior. Overall, GDO provides a goal-driven data optimization framework that enables faster convergence with fewer training samples under a fixed training protocol. Code is available at https://github.com/rujiewu/GDO.

In real-world deployment, vision-language models often encounter disturbances such as weather, occlusion, and camera motion. Under such conditions, their understanding and reasoning degrade substantially, revealing a gap between clean, controlled (i.e., unperturbed) evaluation settings and real-world robustness. To address this limitation, we propose ROVA, a novel training framework that improves robustness by modeling a robustness-aware consistency reward under spatio-temporal corruptions. ROVA introduces a difficulty-aware online training strategy that prioritizes informative samples based on the model's evolving capability. Specifically, it continuously re-estimates sample difficulty via self-reflective evaluation, enabling adaptive training with a robustness-aware consistency reward. We also introduce PVRBench, a new benchmark that injects real-world perturbations into embodied video datasets to assess both accuracy and reasoning quality under realistic disturbances. We evaluate ROVA and baselines on PVRBench, UrbanVideo, and VisBench, where open-source and proprietary models suffer up to 35% and 28% drops in accuracy and reasoning under realistic perturbations. ROVA effectively mitigates performance degradation, boosting relative accuracy by at least 24% and reasoning by over 9% compared with baseline models (QWen2.5/3-VL, InternVL2.5, Embodied-R). These gains transfer to clean standard benchmarks, yielding consistent improvements.

Token reduction accelerates long-video vision--language models (VLMs), but existing methods target Transformers, where reduction is treated as token pruning. We study token reduction in hybrid Mamba--Transformer VLMs and find that it is \emph{stateful}: Mamba layers maintain a recurrent state that accumulates information from earlier tokens, allowing discarded tokens to persist, so reduction behaves more like compression than dropping.We support this view with a representation-based probing method measuring how much information from discarded tokens is retained, and analyze layer-wise sparsity and cross-layer importance stability. Our findings show importance is sparse within layers but unstable across layers, making aggressive early pruning unreliable while hybrids remain robust to later reduction.Motivated by this, we propose a hybrid-aware token reduction framework with a low-to-high progressive schedule and a unified query-conditioned importance score for attention and Mamba layers. For Mamba, excluding the position-dependent decay from the recurrence produces a stronger selection signal. Across long-video benchmarks, our method achieves $3.8{\times}$--$4.2{\times}$ prefilling speedups at a 25% token budget while maintaining near-baseline accuracy and improving with light finetuning. Hybrid models benefit from aggressive reduction, improving both efficiency and accuracy, whereas Transformers exhibit the standard trade-off. Our method also outperforms prior baselines on the same hybrid backbone and combines effectively with visual redundancy reduction methods.

In real scenarios, videos can span several minutes or even hours. However, existing research on spatio-temporal video grounding (STVG), given a textual query, mainly focuses on localizing targets in short videos of tens of seconds, typically less than one minute, which limits real-world applications. In this paper, we explore Long-Form STVG (LF-STVG), which aims to locate targets in long-term videos. Compared with short videos, long-term videos contain much longer temporal spans and more irrelevant information, making it difficult for existing STVG methods that process all frames at once. To address this challenge, we propose an AutoRegressive Transformer architecture for LF-STVG, termed ART-STVG. Unlike conventional STVG methods that require the entire video sequence to make predictions at once, ART-STVG treats the video as streaming input and processes frames sequentially, enabling efficient handling of long videos. To model spatio-temporal context, we design spatial and temporal memory banks and apply them to the decoders. Since memories from different moments are not always relevant to the current frame, we introduce simple yet effective memory selection strategies to provide more relevant information to the decoders, significantly improving performance. Furthermore, instead of parallel spatial and temporal localization, we propose a cascaded spatio-temporal design that connects the spatial decoder to the temporal decoder, allowing fine-grained spatial cues to assist complex temporal localization in long videos. Experiments on newly extended LF-STVG datasets show that ART-STVG significantly outperforms state-of-the-art methods, while achieving competitive performance on conventional short-form STVG. Our code is at: https://github.com/HengLan/ART-STVG.

We present Xray-Visual, a unified vision model architecture for large-scale image and video understanding trained on industry-scale social media data. Our model leverages over 15 billion curated image-text pairs and 10 billion video-hashtag pairs from Facebook and Instagram, employing robust data curation pipelines that incorporate balancing and noise suppression strategies to maximize semantic diversity while minimizing label noise. We introduce a three-stage training pipeline that combines self-supervised MAE, semi-supervised hashtag classification, and CLIP-style contrastive learning to jointly optimize image and video modalities. Our architecture builds on a Vision Transformer backbone enhanced with efficient token reorganization (EViT) for improved computational efficiency. Extensive experiments demonstrate that Xray-Visual achieves state-of-the-art performance across diverse benchmarks, including ImageNet for image classification, Kinetics and HMDB51 for video understanding, and MSCOCO for cross-modal retrieval. The model exhibits strong robustness to domain shift and adversarial perturbations. We further demonstrate that integrating large language models as text encoders (LLM2CLIP) significantly enhances retrieval performance and generalization capabilities, particularly in real-world environments. Xray-Visual establishes new benchmarks for scalable, multimodal vision models, while maintaining superior accuracy and computational efficiency.

Reliable video understanding requires high-quality video datasets that can provide both precise semantic labels and temporally consistent annotations. Detecting annotation errors in densely labeled videos is challenging because errors may arise from semantic mislabeling, where labels disagree with visual content, or temporal disordering, where otherwise plausible labels violate procedural progression. Training dynamics have been used to identify mislabeled training examples primarily for static samples. We investigate checkpoint loss dynamics for out-of-sample auditing of temporally annotated videos. We compute Cumulative Sample Loss (CSL) as the mean annotation-conditioned loss of an audit frame across checkpoints trained on a disjoint reference set. CSL acts as a dynamic fingerprint and captures the persistent disagreement between its annotation and learned visual-temporal structure. High-CSL frames are then flagged as likely candidates for potential annotation errors, including semantic mislabeling or temporal disordering. Experiments on EgoPER and Cholec80 show that CSL substantially outperforms final-checkpoint loss and achieves up to a 4.2-point AUC improvement over prior baselines on EgoPER and 92.0/78.5 AUC for mislabeling/disordering on Cholec80. These results demonstrate checkpoint loss dynamics as an effective diagnostic for temporal annotation auditing.

Video-Language Models (VidLMs) achieve strong benchmark scores, yet these scores often hide whether models use the video at all. We show that VidLM failures follow two pathways: some visual signals are never reliably encoded, while others are encoded but overridden by model priors. We introduce REVEAL, a diagnostic stress-test benchmark for quantifying when and why VidLMs under-use visual evidence. REVEAL contains five controlled probes: camera-motion sensitivity, cross-frame integration, video sycophancy, language-only shortcuts, and temporal expectation bias. Together, they test whether models encode basic video signals, combine evidence across frames, and preserve visual evidence against user assertions, language cues, and learned event expectations. Across 12 VidLMs we find systematic failures along both pathways, with most models falling below chance on the binary and six-way probes that humans solve at 78--100% accuracy. Under assertive prompts, a model's output distribution becomes nearly invariant to whether it is shown a real video or random noise, making visual evidence effectively causally inert. We further carry out mechanistic probes to identify where these failures arise in the model pipeline and why visual evidence is lost. REVEAL provides a scalable, human-verified framework for moving beyond aggregate scores toward structured, reproducible evaluation of multimodal reliability.

Saliency estimation has received growing attention in recent years due to its importance in a wide range of applications. In the context of 360-degree video, it has been particularly valuable for tasks such as viewport prediction and immersive content optimization. In this paper, we propose SalFormer360, a novel saliency estimation model for 360-degree videos built on a transformer-based architecture. Our approach is based on the combination of an existing encoder architecture, SegFormer, and a custom decoder. The SegFormer model was originally developed for 2D segmentation tasks, and it has been fine-tuned to adapt it to 360-degree content. To further enhance prediction accuracy in our model, we incorporated Viewing Center Bias to reflect user attention in 360-degree environments. Extensive experiments on the three largest benchmark datasets for saliency estimation demonstrate that SalFormer360 outperforms existing state-of-the-art methods. In terms of Pearson Correlation Coefficient, our model achieves 8.4% higher performance on Sport360, 2.5% on PVS-HM, and 18.6% on VR-EyeTracking compared to previous state-of-the-art.

In recent years, large language models (LLMs) have made rapid progress in information retrieval, yet existing research has mainly focused on text or static multimodal settings. Open-domain video shot retrieval, which involves richer temporal structure and more complex semantics, still lacks systematic benchmarks and analysis. To fill this gap, we introduce ShotFinder, a benchmark that formalizes editing requirements as keyframe-oriented shot descriptions and introduces five types of controllable single-factor constraints: Temporal order, Color, Visual style, Audio, and Resolution. We curate 1,210 high-quality samples from YouTube across 20 thematic categories, using large models for generation with human verification. Based on the benchmark, we propose ShotFinder, a text-driven three-stage retrieval and localization pipeline: (1) query expansion via video imagination, (2) candidate video retrieval with a search engine, and (3) description-guided shot localization. Experiments on multiple closed-source and open-source models reveal a significant gap to human performance, with clear imbalance across constraints: temporal localization is relatively tractable, while color and visual style remain major challenges. These results reveal that open-domain video shot retrieval is still a critical capability that multimodal large models have yet to overcome.

This paper presents Youtu-Parsing, an efficient and versatile document parsing model designed for high-performance content extraction. The architecture employs a native Vision Transformer (ViT) featuring a dynamic-resolution visual encoder to extract shared document features, coupled with a prompt-guided Youtu-LLM-2B language model for layout analysis and region-prompted decoding. Leveraging this decoupled and feature-reusable framework, we introduce a high-parallelism decoding strategy comprising two core components: token parallelism and query parallelism. The token parallelism strategy concurrently generates up to 64 candidate tokens per inference step, which are subsequently validated through a verification mechanism. This approach yields a 5--11x speedup over traditional autoregressive decoding and is particularly well-suited for highly structured scenarios, such as table recognition. To further exploit the advantages of region-prompted decoding, the query parallelism strategy enables simultaneous content prediction for multiple bounding boxes (up to five), providing an additional 2x acceleration while maintaining output quality equivalent to standard decoding. Youtu-Parsing encompasses a diverse range of document elements, including text, formulas, tables, charts, seals, and hierarchical structures. Furthermore, the model exhibits strong robustness when handling rare characters, multilingual text, and handwritten content. Extensive evaluations demonstrate that Youtu-Parsing achieves state-of-the-art (SOTA) performance on both the OmniDocBench and olmOCR-bench benchmarks. Overall, Youtu-Parsing demonstrates significant experimental value and practical utility for large-scale document intelligence applications.

The advent of always-on personal AI assistants, enabled by all-day wearable devices such as smart glasses, demands a new level of contextual understanding, one that goes beyond short, isolated events to encompass the continuous, longitudinal stream of egocentric video. Achieving this vision requires advances in long-horizon video understanding, where systems must interpret and recall visual and audio information spanning days or even weeks. Existing methods, including large language models and retrieval-augmented generation, are constrained by limited context windows and lack the ability to perform compositional, multi-hop reasoning over very long video streams. In this work, we address these challenges through EGAgent, an enhanced agentic framework centered on entity scene graphs, which represent people, places, objects, and their relationships over time. Our system equips a planning agent with tools for structured search and reasoning over these graphs, as well as hybrid visual and audio search capabilities, enabling detailed, cross-modal, and temporally coherent reasoning. Experiments on the EgoLifeQA and Video-MME (Long) datasets show that our method achieves state-of-the-art performance on EgoLifeQA (57.5%) and competitive performance on Video-MME (Long) (74.1%) for complex longitudinal video understanding tasks. Code is available at https://github.com/facebookresearch/egagent.

Despite the rapid progress of video generation models, the role of data in influencing motion is poorly understood. We present Motive (MOTIon attribution for Video gEneration), a motion-centric, gradient-based data attribution framework that scales to modern, large, high-quality video datasets and models. We use this to study which fine-tuning clips improve or degrade temporal dynamics. Motive isolates temporal dynamics from static appearance via motion-weighted loss masks, yielding efficient and scalable motion-specific influence computation. On text-to-video models, Motive identifies clips that strongly affect motion and guides data curation that improves temporal consistency and physical plausibility. With Motive-selected high-influence data, our method improves both motion smoothness and dynamic degree on VBench, achieving a 74.1% human preference win rate compared with the pretrained base model. To our knowledge, this is the first framework to attribute motion rather than visual appearance in video generative models and to use it to curate fine-tuning data.

Semantic Multi-Object Tracking (SMOT) is evolving from purely geometric localization toward comprehensive video understanding. However, existing paradigms predominantly rely on closed-set interaction tags and fragmented perception pipelines, creating a bottleneck that prevents the full utilization of Multi-modal Large Language Models (MLLMs) for dynamic scenes. In this paper, we elevate SMOT from rigid classification to an open-ended generative reasoning task. To support this paradigm shift, we introduce Grand-SMOT, a large-scale benchmark featuring high-density, dual-stream narratives. This dataset explicitly decouples micro-level individual dynamics from macro-level environmental contexts, directly resolving the semantic scarcity of prior tracking datasets. Furthermore, we propose LLMTrack, a unified MLLM-driven framework for dynamic SMOT. Guided by a verifiable ``\textit{Macro-Understanding-First}'' mechanism, LLMTrack employs a Spatio-Temporal Fusion Module to compress discrete geometric trajectories into continuous semantic tokens, effectively suppressing temporal hallucinations in long-sequence tracking. Extensive experiments, utilizing a novel decoupled evaluation protocol, validate that LLMTrack achieves state-of-the-art geometric tracking robustness while delivering a qualitative leap in generative semantic reasoning. The code and datasets are publicly available at https://github.com/liaopan-lp/LLMTrack-GrandSMOT.

Transferring articulated motion from monocular videos to rigged 3D characters is challenging due to pose ambiguity in 2D observations and morphological differences between source and target. Existing approaches often follow a reconstruct-then-retarget paradigm, tying transfer quality to intermediate 3D reconstruction and limiting applicability to categories with parametric templates. We propose MorphGS, a framework that formulates motion retargeting as a target-driven analysis-by-synthesis problem, directly optimizing target morphology and pose through image-space supervision. A rig-coupled morphology parameterization factorizes character identity from time-varying joint rotations, while dense 2D-3D correspondences and synthesized views provide complementary structural and multi-view guidance. Experiments on synthetic benchmarks and real-world videos show consistent improvements over baselines. Project page: https://xodus777.github.io/MorphGS/

Intrinsic image decomposition aims to recover material and illumination factors from RGB observations, but real-world images entangle reflectance with illumination, visibility, shadows, and non-diffuse appearance. Recent single-image methods address this entanglement with a residual image formation model, decomposing RGB into albedo, diffuse shading, and a non-diffuse residual. However, applying such decomposition independently to consecutive frames lacks scene-level context, leading to inconsistent factor assignments, albedo drift, and leakage across views. Meanwhile, multi-view inverse-rendering methods recover properties, such as material, lighting, and geometry, but they rely on synthetic data due to the highly uncertain estimation, and do not generalize to supervision by real images. We present MVID, Multi-View Intrinsic image Decomposition, a feed-forward framework built on the residual image formation model. MVID builds a scene-level multi-view representation and decodes a view-consistent albedo together with coherent per-view shading and residual factors through a factor query adapter, while using the same image formation model for self-supervised RGB reconstruction on unlabeled real-world sequences. Experiments on indoor, real-world, and outdoor benchmarks show that MVID improves both per-frame decomposition quality and cross-view consistency over single-view intrinsic, generative intrinsic, and multi-view inverse-rendering baselines. The resulting view-stable factors support practical image-space applications, including multi-view consistent illumination editing and specularity removal.

Short-form video moderation is increasingly pushed toward edge and privacy-sensitive settings, where users may intend videos for a limited audience, such as friends or private groups, but sending raw clips to a central server can broaden exposure, consume bandwidth, and add moderation latency. Federated learning can keep videos on device, but unprotected model updates may still leak information, and full-video backbones are expensive to communicate. We present FedVideoMAE, a privacy-preserving federated framework for violence detection that adapts a frozen VideoMAE backbone with lightweight LoRA and prompt parameters. Each training round combines self-supervised masked video reconstruction with client-side differential privacy and pairwise masked aggregation (SA) of adapter updates. Violence labels are held out from federation and used only for downstream evaluation, separating private representation learning from supervised assessment. On RWF-2000, exchanging 5,518,848 trainable parameters instead of the 156,371,328-parameter instantiated pretraining state gives a 28.3x model-state payload ratio. FedVideoMAE reaches 77.25% test accuracy without DP or SA, while accuracy under DP+SA remains in the 65.25-66.00% range. Transfer experiments on RLVS and binary UCF-Crime show similar behavior. These results characterize the privacy-utility trade-off for edge video moderation. Code is available at: https://github.com/zyt-599/FedVideoMAE

We present TCAM (Track and Caption Any Motion), a generative framework that watches a video and with no text query and no region prompt decides what is moving, describes each motion in open vocabulary, locates it in time, and points to the exact trajectories that carry it. Two mature lines of work make this possible yet leave it unsolved: dense point trackers follow pixels with sub-object precision but emit no language, while video-language models produce fluent descriptions only when handed a query and only from clip-level features that cannot resolve which pixels move. Object-level captioners narrow the gap but still reason over detector boxes or masks, never reaching individual trajectories. TCAM couples tracking and language at point granularity through a Caption-Aware Resampler, where a small set of learnable queries cross-attends to dense point trajectory tokens and distills them into a fixed-length motion context that conditions a language decoder. The decoder generates an entire video's events in a single pass, each with a free-form caption, a start and end time, and a pointer to the trajectories it refers to, for sequential events and several subjects active at once. Training uses only existing segmentation annotations, with no extra event labeling, to supervise caption quality, pointer-mask alignment, and pointer diversity. On over 50K clips, TCAM outperforms dense video captioning baselines and matches dedicated, query-based grounding and point-tracking methods despite using no query, showing that trajectory-conditioned generation is a direct route to motion-driven video understanding.

Despite rapid advances in video generative models, robust metrics for evaluating visual and temporal correctness of complex human actions remain elusive. Critically, existing pure-vision encoders and Multimodal Large Language Models (MLLMs) are strongly appearance-biased, lack temporal understanding, and thus struggle to discern intricate motion dynamics and anatomical implausibilities in generated videos. We tackle this gap by introducing a novel evaluation metric derived from a learned latent space of real-world human actions. Our method first captures the nuances, constraints, and temporal smoothness of real-world motion by fusing appearance-agnostic human skeletal geometry features with appearance-based features. We posit that this combined feature space provides a robust representation of action plausibility. Given a generated video, our metric quantifies its action quality by measuring the distance between its underlying representations and this learned real-world action distribution. For rigorous validation, we develop a new multi-faceted benchmark specifically designed to probe temporally challenging aspects of human action fidelity. Through extensive experiments, we show that our metric achieves substantial improvement of more than 68% compared to existing state-of-the-art methods on our benchmark, performs competitively on established external benchmarks, and has a stronger correlation with human perception. Our in-depth analysis reveals critical limitations in current video generative models and establishes a new standard for advanced research in video generation.

While Multimodal Large Language Models (MLLMs) have become adept at recognizing objects, they often lack the intuitive, human-like understanding of the world's underlying physical and social principles. This high-level vision-grounded semantics, which we term visual knowledge, forms a bridge between perception and reasoning, yet remains an underexplored area in current MLLMs. To systematically evaluate this capability, we present VKnowU, a comprehensive benchmark featuring 1,680 questions in 1,249 videos, covering 8 core types of visual knowledge spanning both world-centric (e.g., intuitive physics) and human-centric (e.g., subjective intentions). Evaluation of 28 SOTA MLLMs reveals that leading models still fall short of human performance, with particularly notable gaps in the world-centric. To bridge this gap, we introduce a new dataset, VKnowQA, and VideoKnow+, a baseline model that explicitly incorporates visual knowledge into MLLMs. VideoKnow+ follows a structured See-Think-Answer paradigm and adopts reinforcement learning with visual knowledge reward, achieving a +3.7% improvement on VKnowU and consistent gains on MVBench (+5.4%), Video-MME (+7.0%), and MMVU (+5.7%). Our work highlights visual knowledge as a missing cornerstone for developing more generalizable MLLMs that can not only see but also truly understand our worlds.

Unmanned aerial vehicles (UAVs) have become powerful platforms for real-time, high-resolution data collection, producing massive volumes of aerial videos. Efficient retrieval of relevant content from these videos is crucial for applications in urban management, emergency response, security, and disaster relief. While text-video retrieval has advanced in natural video domains, the UAV domain remains underexplored due to limitations in existing datasets, such as coarse and redundant captions. Thus, in this work, we construct the Drone Video-Text Match Dataset (DVTMD), which contains 2,864 videos and 14,320 fine-grained, semantically diverse captions. The annotations capture multiple complementary aspects, including human actions, objects, background settings, environmental conditions, and visual style, thereby enhancing text-video correspondence and reducing redundancy. Building on this dataset, we propose the Text-Conditioned Multi-granularity Alignment (TCMA) framework, which integrates global video-sentence alignment, sentence-guided frame aggregation, and word-guided patch alignment. To further refine local alignment, we design a Word and Patch Selection module that filters irrelevant content, as well as a Text-Adaptive Dynamic Temperature Mechanism that adapts attention sharpness to text type. Extensive experiments on DVTMD and CapERA establish the first complete benchmark for drone text-video retrieval. Our TCMA achieves state-of-the-art performance, including 45.5% R@1 in text-to-video and 42.8% R@1 in video-to-text retrieval, demonstrating the effectiveness of our dataset and method. The code and dataset will be released.

Unified multimodal models have shown promising results in multimodal content generation and editing but remain largely limited to the image domain. In this work, we present UniVideo, a versatile framework that extends unified modeling to the video domain. UniVideo adopts a dual-stream design, combining a Multimodal Large Language Model (MLLM) for instruction understanding with a Multimodal DiT (MMDiT) for video generation. This design preserves the MLLM's original text generation capabilities, enables accurate interpretation of complex multimodal instructions, and maintains visual consistency in the generated content. Built on this architecture, UniVideo unifies diverse video generation and editing tasks under a single multimodal instruction paradigm and is jointly trained across them. Extensive experiments demonstrate that UniVideo matches or surpasses state-of-the-art task-specific baselines in text/image-to-video generation, in-context video generation and in-context video editing. Notably, the unified design of UniVideo enables two forms of generalization. First, UniVideo supports task composition, such as combining editing with style transfer, by integrating multiple capabilities within a single instruction. Second, even without explicit training on free-form video editing, UniVideo transfers its editing capability from large-scale image editing data to this setting, handling unseen instructions such as changing the environment or altering materials within a video. Beyond these core capabilities, UniVideo also supports visual-prompt-based video generation, where the MLLM interprets visual prompts and guides the MMDiT during synthesis. To foster future research, we released our model and code.

Recent advances in large video-language models (VLMs) rely on extensive fine-tuning techniques that strengthen alignment between textual and visual comprehension. Leading pipelines typically pair supervised fine-tuning (SFT) with reinforcement learning from preference data to enhance video comprehension. However, as VLMs scale in parameter size, so does the cost of gathering enough human feedback. To make fine-tuning more cost-effective, recent frameworks explore reinforcement learning with AI feedback (RLAIF), which replace human preference with AI as a judge. Current RLAIF frameworks rely on a specialized reward model trained with video narratives to create calibrated scalar rewards -- an expensive and restrictive pipeline. We propose Oracle-RLAIF, a novel framework that replaces the trained reward model with a more general Oracle ranker which acts as a drop-in model ranking candidate model responses rather than scoring them. Alongside Oracle-RLAIF, we introduce $GRPO_{rank}$, a novel rank-based loss function based on Group Relative Policy Optimization (GRPO) that directly optimizes ordinal feedback with rank-aware advantages. Empirically, we demonstrate that Oracle-RLAIF consistently outperforms leading VLMs using existing fine-tuning methods when evaluated across various video comprehension benchmarks. Oracle-RLAIF paves the path to creating flexible and data-efficient frameworks for aligning large multi-modal video models with reinforcement learning from rank rather than score.

Multimodal Large Language Models (MLLMs) often struggle with fine-grained perception, such as identifying small objects in high-resolution images or detecting key moments in long videos. Existing methods typically rely on complex, task-specific fine-tuning, which reduces generalizability and increases system complexity. In this work, we propose an effective, training-free framework that uses an MLLM's intrinsic uncertainty as proactive guidance. Our core insight is that a model's uncertainty decreases when provided with relevant visual information. We introduce a unified mechanism that scores candidate visual inputs by response uncertainty, enabling the model to autonomously focus on the most informative data. We apply this simple principle to three challenging visual tasks: Visual Search, Long Video Understanding, and Temporal Grounding, allowing off-the-shelf MLLMs to achieve performance competitive with specialized, fine-tuned systems. Our results demonstrate that leveraging intrinsic uncertainty is a powerful strategy for improving fine-grained multimodal performance.

Existing Causal-Why Video Question Answering (VideoQA) models often struggle with higher-order reasoning, relying on opaque, monolithic pipelines that entangle video understanding, causal inference, and answer generation. These black-box approaches offer limited interpretability and tend to depend on shallow heuristics. We propose a novel, modular paradigm that explicitly decouples causal reasoning from answer generation, introducing natural language causal chains as interpretable intermediate representations. Inspired by human cognitive models, these structured cause-effect sequences bridge low-level video content with high-level causal reasoning, enabling transparent and logically coherent inference. Our two-stage architecture comprises a Causal Chain Extractor (CCE) that generates causal chains from video-question pairs, and a Causal Chain-Driven Answerer (CCDA) that derives answers grounded in these chains. To address the lack of annotated reasoning traces, we introduce a scalable method for generating accurate causal chains from existing datasets. We construct human verified causal chains for 46K samples. We also propose CauCo, a new evaluation metric for causality-oriented captioning. Experiments on three large-scale benchmarks demonstrate that our approach not only outperforms state-of-the-art models, but also yields substantial gains in explainability, user trust, and generalization -- positioning the CCE as a reusable causal reasoning engine across diverse domains. Project page: https://paritoshparmar.github.io/chainreaction/

Text-to-video (T2V) models now produce striking visuals, yet they routinely violate everyday physics; objects float, tools are misused, and causal sequences break down. Existing benchmarks mostly probe isolated physical laws and rely on vision-language models to score videos directly, which entangles perception and reasoning in one judgment and correlates poorly with humans. We introduce PhysVidBench, a benchmark of 383 base prompts, expanded to 766 prompts with enriched variants and 4,123 manually reviewed QA items, together with a language-grounded evaluation framework that takes a different route: instead of asking a VLM "Is this video physically correct?", we caption the video, then ask a language model to answer prompt-derived yes/no physics questions using only the captions. This split between seeing and reasoning makes each score traceable to the captions that justify it and aligns more closely with human judgment than direct VLM scoring (Pearson r=0.45-0.69, compared with 0.21-0.47 for the strongest direct VLM evaluator). To test the framework, we carefully curate a set of human-validated prompts spanning seven physical dimensions, with a focus on tool use and affordances - areas largely absent from prior physics-focused benchmarks. Across 12 open and proprietary T2V systems, the best model reaches only 36.2% average accuracy, and no model consistently handles everyday physical reasoning. The same pipeline can also guide iterative error-guided prompt refinement, improving CogVideoX-2B from 21.6 to 32.7 and CogVideoX-5B from 17.8 to 29.7 without retraining.

While CNNs and ViTs dominate vision architectures, all-MLP models offer a structurally simpler alternative whose patch-independent processing is naturally suited to exploiting temporal redundancy in video. We present evMLP, an all-MLP architecture that processes image patches independently, enabling an event-driven local update mechanism for video processing: by defining inter-frame changes as "events" and processing only the patches where events occur, evMLP avoids redundant computation on unchanged regions. Because each patch is processed independently, skipping an unchanged patch leaves all other outputs unaffected; at an event threshold of zero, the mechanism produces outputs identical to the dense baseline rather than an approximation. On ImageNet, evMLP achieves 73.5% top-1 accuracy at 1.03 GMACs (rising to 77.0% with knowledge distillation and an extended training schedule). On multiple video datasets, the event-driven mechanism reduces computational cost by 8.4%-26.8% while maintaining output consistency with the dense baseline. Wall-clock measurements confirm that these savings translate into actual speedup under compute-bound conditions, and that stream-level parallelism is the effective deployment strategy for multi-core systems. The code and pre-trained models are available at https://github.com/i-evi/evMLP.

We propose a step-by-step video-to-audio (V2A) generation method that provides finer control over the generation process and more realistic audio synthesis. Inspired by traditional Foley workflows, our approach enables incremental generation of complementary sounds, allowing users to author multiple sound events induced by a video. To avoid the need for costly multi-reference video-audio datasets, each generation step is formulated as a negatively guided V2A process that discourages duplication of sounds already present in previously generated tracks. The guidance model is trained by finetuning a pre-trained V2A model on audio pairs from non-overlapping segments of the same video, encouraging it to leverage acoustic context while remaining visually grounded, and enabling training with standard single-reference audiovisual datasets. Objective and subjective evaluations demonstrate that our method enhances the separability of generated sounds at each step and improves the overall quality of the final composite audio, outperforming existing baselines. Our project page is available at: https://ahykw.github.io/sbsv2a/.

Motion blur remains one of the most common and visually disruptive degradations in real-world smartphone imaging, yet existing deblurring benchmarks are often limited in scale, resolution, or domain relevance. This gap is especially pronounced for smartphones, where rolling shutter, small sensors, and ISP processing produce blur statistics that differ from GoPro/DSLR-based benchmarks. We introduce a large-scale smartphone-oriented deblurring dataset constructed from 240~fps slow-motion video. To approximate exposure-time radiance integration, we synthesize blur by temporally averaging a fixed window of $N=30$ consecutive frames, which corresponds to an effective exposure of $T=1/8$~second, and we select the temporally centered frame as the sharp ground truth. The resulting benchmark contains 42,045 paired blur--sharp images at $1920\times1080$ resolution spanning 843 distinct scenes, with a train/test split of 37,841/4,204 pairs. We benchmark multiple state-of-the-art deblurring models using PSNR and SSIM and observe consistent performance degradation relative to the baseline similarity between the input blurry images and ground truth, underscoring the realism and difficulty of the proposed data. We release the dataset and generation scripts via HuggingFace to facilitate the development and evaluation of robust, deployment-oriented deblurring methods.

The rapid proliferation of surveillance cameras has increased the demand for automated violence detection. While CNNs and Transformers have shown success in extracting spatio-temporal features, they struggle with long-term dependencies and computational efficiency. We propose FuseMamba-VD: Dual Branch VideoMamba with Gated Class Token Fusion (GCTF), an efficient architecture combining a dual-branch design and a state-space model (SSM) backbone where one branch captures spatial features, while the other focuses on temporal dynamics. The model performs continuous fusion via a gating mechanism from the spatial branch into the temporal branch to enhance detection of violent activities even in challenging surveillance scenarios. We also present a new benchmark by merging RWF-2000, RLVS, SURV and VioPeru datasets in video violence detection, ensuring strict separation between training and testing sets. Experimental results demonstrate that our model achieves state-of-the-art performance on this benchmark and also on DVD dataset which is a recently introduced dataset on video violence detection, offering an optimal balance between accuracy and computational efficiency, demonstrating the promise of SSMs for scalable, resource efficient video violence detection. The code and pre-trained models are available at https://github.com/damith92/FuseMamba-VD.