video analysis teams

Abstract: We introduce The Prestige, a video understanding benchmark that probes cognitive visual reasoning through professional magic performances, where deliberate misdirection challenges models to track events and explain unexpected outcomes. The benchmark comprises open-ended questions ranging from temporal and causal understanding to robustness against adversarial questions. Evaluation of leading open-source and proprietary VLMs reveals a substantial gap between even the strongest model and human performance. We also find systematic discrepancies between human and automated grading and identify recurring model failure modes. Ablations further show that models can select correct multiple-choice answers even when they fail to generate correct explanations independently, motivating further investigations into question formats. Together, these findings highlight limitations in both VLM reasoning and its evaluation, underscoring the need for more reliable video understanding models and more rigorous benchmarks.

Abstract: Point-supervised temporal action localization (PTAL) aims to recognize and localize action instances in untrimmed videos using only point-level annotations. Existing pseudo-label-based methods have made substantial progress in PTAL. However, confidence-based initialization can overlook well-localized proposals. Meanwhile, existing methods either keep pseudo-labels fixed during detector training or update them without fully exploiting alternative boundary evidence from the original proposal pool. These limitations constrain supervision quality, thereby hindering detection performance. To learn high-quality pseudo-labels, we propose a framework that integrates prior-guided initialization with consensus-guided refinement during detector training. Our framework comprises two modules: Prior-Guided Initialization (PGI) and Consensus-Guided Refinement (CGR). PGI combines proposal confidence with point-derived priors to establish instance-specific reliability for initial pseudo-label selection and candidate retention. CGR continually uses the retained original candidates to construct boundary targets, combining their consensus with evolving detector predictions to refine pseudo-labels throughout training. Experiments on four benchmarks demonstrate that our method achieves competitive detection performance compared with existing methods. In particular, our method achieves 68.5\% average mAP over tIoU thresholds 0.1--0.7 on THUMOS14, surpassing PseR by 3.7 percentage points. Code is available at https://github.com/zhenyingfang/LHQP

Abstract: Long-video understanding with multimodal LLMs is fundamentally constrained by the mismatch between video length and the model's limited visual context budget, making frame selection essential for practical reasoning. Existing query-aware methods typically rely on heuristic pipelines: they assign one-dimensional query-conditioned scores to frames and then apply sampling, ranking, or extraction rules to form the final subset. While effective in practice, such methods are generally not derived from an explicit optimization objective and struggle to capture the heterogeneous evidence required for complex reasoning. In this paper, we revisit long-video frame selection from a principled optimization perspective. We formulate it as a structured evidence-allocation problem and propose OTFS, an optimal transport framework that maps multiple evidence sources onto video frames. To reflect the asymmetry of this problem---important evidence should be preserved, whereas redundant frames need not absorb mass---we instantiate this view with a semi-unbalanced entropic optimal transport objective and an efficient solver, OTFS-Sinkhorn. Combined with an exact dynamic program for coverage-aware subset extraction, OTFS yields a practical two-stage training-free pipeline. We further show that, in the single-source case, our formulation induces a Gibbs-type distribution over frames, making Q-Frame's temperature-scaled distribution a special case of our framework while also providing a broader perspective for understanding related prior methods. Experiments on three long-video understanding benchmarks show that OTFS consistently outperforms strong training-free baselines. Codes are available at https://github.com/barry664/OTFS.

Abstract: Recent Omni LLMs (OLLMs) have demonstrated remarkable capabilities in joint audio-visual reasoning, yet cross-modal hallucinations remain a major challenge. Existing mitigation methods mainly operate at the output level, overlooking the entanglement of grounding evidence and learned priors within fused representations. To address this issue, we propose Omni-Woodpecker (OWP), a training-free framework that shifts hallucination mitigation from output-level manipulation to conflict-conditioned representation rebalancing. Specifically, OWP first identifies cross-modal conflicts from complementary modality-specific views. It then employs orthogonal decomposition to separate grounding evidence from competing priors within the hidden states. Finally, OWP adaptively rebalances fused representations by suppressing misleading priors and restoring ignored evidence before decoding. Extensive experiments demonstrate the effectiveness and strong cross-architecture generalization of OWP. Using Qwen2.5-Omni-7B as the backbone, OWP improves performance by 12.9 and 8.6 points on CMM and AVHBench, respectively, while consistently achieving gains of 0.67-2.25 points across four general OLLM reasoning benchmarks. Furthermore, OWP generalizes effectively to VideoLLaMA2-AV, improving performance by 4.6 and 3.9 points on the same benchmarks. Code is publicly available at https://github.com/ZiiJianLiu/Omni-Woodpecker

Abstract: Vision-language model (VLM) inference incurs substantial prefill latency and memory overhead due to the long visual token sequences of large visual inputs. Existing serving systems typically exploit exact cache reuse through prefix caching or position-independent caching, overlooking the potential of approximate caching in real-world workloads. However, directly reusing an approximate cache entry may introduce spurious visual information, which naive approaches that select a subset of important visual tokens for recomputation fail to adequately mitigate this risk. We present VisMemo, an efficient serving system that enables approximate KV cache reuse for modern VLMs. Its core is a system-algorithm co-design that combines an entropy-aware KV cache reuse algorithm with differentiated cache management. Specifically, VisMemo uses patch entropy to identify reuse candidates and warmup attention to prioritize recomputation based on the current query. To support the large cache capacity across GPU memory, CPU memory, and SSD storage, VisMemo applies distinct I/O and cache policies to exact and approximate hits: sparse patch-level I/O for approximate hits, prioritized retention for exact hits, and a reserved CPU victim cache for preset entries selected using offline workload priors and online reuse observations. VisMemo further optimizes the memory access path between the inference engine and the storage backend, eliminating redundant caching and unnecessary data movement. Experiments on multiple public datasets show that VisMemo preserves output quality under approximate reuse. On a real-world production trace, VisMemo increases the overall cache hit rate from 27.78% to 61.45% and reduces prefill latency by up to 48.3% at 8K input tokens.

Abstract: As Vision-Language Models (VLMs) advance, they are increasingly expected to perform spatiotemporal reasoning over motion, actions, and view changes beyond what the input images show. Video World Models (VWMs) provide a natural unified engine for visual imagination by generating task-relevant visual content as additional evidence. However, existing approaches typically invoke a VWM through fixed pipelines centered on a predefined form of visual imagination, such as novel-view synthesis. To support flexible visual imagination for diverse reasoning tasks, we introduce CoWorld Reasoner, an agentic framework that uses a VWM as a unified visual reasoning engine. Its Video World Model-Driven Harness exposes multiple visual operations through tools and uses skills to guide the VLM in deciding whether to imagine, selecting an operation, constructing a generation request, and interpreting the resulting visual evidence. We further introduce VLM-VWM Harness Co-Evolution, which jointly revises tools, skills, and execution code based on execution feedback while keeping both model weights fixed. Across seven benchmarks spanning spatial intelligence, temporal and causal reasoning, and perceptual reasoning, CoWorld Reasoner outperforms direct reasoning with the same VLM. The strongest variant improves on all seven benchmarks, including relative accuracy gains of 11.8% on MMSI and 13.0% on ERQA, and outperforms all five evolving baselines using the same backbone across five benchmarks.

Abstract: Large multimodal models (LMMs) need to understand how entities move and interact as a scene develops. This requires following the same entities over time and relating their observed motion to subsequent changes. We formulate Scene Evolution Understanding (SEU), a task for predicting and interpreting this change. Given observed video, a model estimates entity histories and predicts one scene continuation through future trajectories and Events, accompanied by Visual Evidence from observed motion and a scene description. We construct SceneForesee, a benchmark for evaluating entity continuity and scene evolution reasoning through paired observed and future scenes with shared entity identities. It contains 144.57K temporal pairs, 1.05M entity correspondences, and 272.49K Events paired with observed evidence. We propose a spatiotemporal reasoning framework, EvoSee, that learns future geometry and Event semantics through shared candidate states and joint assignment. It selects one continuation per entity from observation and uses the selected changes for entity refinement, retention, and scene representation. On SceneForesee, EvoSee outperforms Qwen3.5-9B by 10.45 points in Combined Point-F1 and 17.24 points in Event Global F1. On the 4D reasoning benchmark Dyn-Bench, it improves question answering accuracy and grounding J&F by 4.8 and 3.58 points over their respective base systems. These results support scene evolution as both a prediction target and a dynamic representation that brings predicted entity changes into spatiotemporal reasoning.The project is available at https://github.com/mikubaka88/SceneForesee-Spatiotemporal-Reasoning.

Abstract: To adapt billion-parameter Vision Transformers for long-video understanding, recent methods freeze the backbone and train lightweight convolutional modules. While effective for parameter-efficient training, existing adapters do not reduce inference-time computation, leaving scalability with respect to video length largely unaddressed. Token selection can reduce attention cost by pruning redundant tokens, but it breaks the spatial grid structure required by convolutional adapters. This forces an expensive dense reconstruction, nullifying much of the potential speedup. We address this by introducing native sparse 2D convolutions, a primitive that allows these adapters, for the first time, to operate directly and efficiently on dynamically pruned token sets. We integrate this primitive into SV-TAD, an adapter framework for temporal action detection, reducing VideoMAEv2-L computation by up to 64% and achieving 2.2x faster inference, while maintaining state-of-the-art accuracy on THUMOS-14 and ActivityNet-1.3. When scaled to InternVideoNext-L, our approach surpasses the previous state of the art at roughly half its computational cost. Moreover, the sparse formulation naturally supports auxiliary task tokens, which improves fine-grained assembly detection on ATTACH.

Abstract: Streaming video LLMs must retain evidence before its relevance to future tasks is known and respond when sufficient evidence becomes available. The challenge is to form reusable factual memory without compromising real-time perception. We introduce OneStreamer, which jointly learns query-independent evidence recording and task response through a shared proactive generation process. Its Proactive Hierarchical Caption Memory (PHCM) produces time-grounded local-detail captions and summaries of completed events. Streaming caption targets supervise the interpretation of observed video prefixes during training. At inference, model-generated records complement a recent visual window, providing reusable factual context without revisiting historical visual features. Proactive State Transition Learning (PSTL) reduces the dominance of repeated waiting states by preserving supervision at all output anchors and selecting representative state-change and state-persistence tokens. We further develop a streaming data synthesis pipeline that aligns output content and timing with available evidence. Combining the resulting streaming captions and QA with cleaned open-source data yields OneStreamer-1M, a broad-coverage streaming video interaction dataset with over one million records spanning diverse tasks. Our 4B model achieves the best results among the compared methods across all eight evaluated streaming video understanding benchmarks. Ablations show that retaining generated captions improves historical QA without degrading real-time perception. PSTL also outperforms dense state supervision while supervising only 27.5% of annotated state tokens. Together, these results support proactive generation as a shared learning interface connecting perception, memory formation, and timely response in streaming video interaction.

Abstract: Long video understanding relies on video memory to overcome the context limits of multimodal large language models. Existing methods follow a build-then-reasoning pipeline: memory is built offline for the entire video, then reasoned over as a static source. In practice a long video is shared by several questions, and this pipeline is costly at both ends: with few questions, building memory for the whole video costs far more than answering them; with many questions, the memory is never updated, so what is learned while answering questions is lost to the next question. To alleviate these, we introduce Sprout, an agentic framework that builds memory while reasoning: a temporal tree that sprouts detailed nodes as questions are answered. The agent watches the video segment by segment at a low frame rate, stopping when the current question can be answered, remembers each segment as a coarse node of the tree, and revisits key intervals at a higher frame rate to refine the tree with the recovered details. Once a segment is recorded as text, its video input is removed from the context history, while the original video remains reachable through the video tools. The memory tree and prior question--answer records persist across questions, so the memory is online and dynamic: built from the first question onward and updated by every question thereafter. We find that replacing accumulated video inputs with textual memory substantially reduces context usage while maintaining accuracy, with slight improvements in some settings. Across benchmarks on three models, Sprout achieves competitive or improved accuracy relative to representative offline memory methods, with no upfront construction stage and lower context cost per question.

Abstract: Recent omni-modal models demonstrate strong perception of audio and visual inputs, yet often struggle to connect what they hear with what they see at the same moment. This weakness in temporal correspondence can cause models to associate spoken cues with the wrong visual scenes, producing plausible answers grounded in incorrect audio-visual pairings. We diagnose this problem through controlled temporal swaps, revealing that model answers do not reliably follow changes in these pairings. To address it, we propose Synchrony-Guided Representation Alignment (SyncRA), a lightweight method for strengthening temporal correspondence between audio and vision. Specifically, SyncRA contrasts intermediate audio-visual representations within each video, aligning matching moments while separating mismatched ones to capture local temporal correspondence within a shared global context. The objective derives supervision directly from existing input timing, requiring no additional annotations and leaving inference unchanged. We evaluate SyncRA across four open omni-modal models spanning different sizes and architectures on five public video benchmarks. SyncRA consistently outperforms answer-only fine-tuning across all model-benchmark combinations, while substantially improving the ability to track changing audio-visual pairings in controlled evaluations. These results demonstrate that lightweight, targeted supervision can effectively strengthen temporal correspondence and translate into broad improvements in audio-visual question answering.

Abstract: Sparse video understanding often requires selecting a small set of visual evidence under a fixed frame budget. Most sparse selectors allocate this budget globally, allowing all frames to compete with one another. For temporally localized events, this can be a poor inductive bias: useful evidence is often distributed across pre-event context, the event itself, and post-event consequences. We study fixed-budget evidence allocation for localized event videos and show that globally competitive Top-$K$ selectors can preserve recognition and grounding while producing unstable event evidence. On DoTA Video Anomaly Recognition, Global Top-$K$ obtains competitive recognition and temporal grounding, but low selector-event alignment at $K=12$ (Frame AUC $51.5 \pm 10.8$). We propose dKFD, a phase-structured differentiable selector that reserves evidence capacity across pre-event, event, and post-event phases after full-sequence temporal encoding. Under matched-budget multi-seed evaluation, dKFD improves Frame AUC by $+30.97$ over a matched Global Top-$K$ selector at $K=12$ ($p<0.01$), while yielding modest but statistically significant recognition gains and comparable temporal grounding. Mechanism ablations show that phase supervision is load-bearing: removing it reduces Frame AUC to $41.1 \pm 12.0$ even when phase-partitioned budgets are retained. Downstream diagnostics on VRU-Accident show consistent gains over learned Global Top-$K$ across VLM families, while dense captioning reveals a boundary condition where uniform sampling remains competitive. These results support phase-structured allocation as a controlled fixed-budget approach for event-centric sparse evidence selection, not as a universal video summarization strategy.

Abstract: Multimodal Large Language Models (MLLMs) have achieved remarkable progress on short video understanding yet remain limited on long videos due to the limited visual context window. Prevailing approaches rely on uniform frame sampling or recent coarse-to-fine agentic zooming, both of which struggle to localize sparse, decisive evidence in sufficiently long videos. We formulate long video understanding as a \textbf{Sequential Evidence Acquisition (SEA)} problem, in which an agent reads the video turn by turn along the temporal axis, deciding at each turn how fast to watch, what evidence to retain, when to revisit uncertain segments, and when to stop and answer. Inspired by this view, we propose \textbf{VideoScout}, a multi-turn reasoning agent that instantiates the SEA paradigm through adaptive reasoning pacing. Specifically, by dynamically controlling the viewing pace, VideoScout enables efficient traversal of long videos within a bounded visual context window, allowing the agent to access more video content while balancing content analysis depth with reading efficiency. To train VideoScout, we construct VideoScout-66K, a set of over 66K high-quality exploration turns from 10K answer-verified trajectories, and adopt a two-stage pipeline: cold-start supervised fine-tuning teaches the agent per-turn output format, while the Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO) algorithm performs trajectory-level reinforcement learning with a composite reward that jointly considers answer accuracy, output format compliance, and the temporal alignment between the agent's viewing progress and the teacher's answer timing measured by intersection-over-union (IoU). Extensive experiments on long video understanding and reasoning benchmarks demonstrate that our 7B model achieves strong performance compared with existing trained 7B agentic models.

Abstract: Long-video understanding remains challenging for multimodal large language models (MLLMs) because densely encoding long frame sequences is computationally expensive, while uniform sampling under a limited visual budget can miss sparse yet decisive evidence. Recent training-free keyframe selection methods have enabled more efficient inference and yielded promising performance gains. However, many existing methods score frames largely in isolation without explicitly considering how each candidate complements the currently selected subset, potentially resulting in redundant selections and incomplete evidence coverage. To address this limitation, we propose MarKey, a training-free framework that formulates keyframe selection as subset-aware greedy optimization. At each iteration, MarKey scores each candidate using a tractable surrogate that jointly accounts for query relevance, marginal coverage gain, and context-dependent redundancy, and selects the frame with the highest utility. To make this iterative subset-aware evaluation efficient, MarKey uses a compact set of representative anchors to approximate full-video coverage and a bounded window of previously selected frames to limit context-dependent comparisons. Experiments on six benchmarks spanning holistic video understanding, human-centric video understanding, and open-ended video understanding demonstrate that MarKey consistently outperforms existing methods. Further analyses show robust gains across different MLLM backbones, model scales, and frame budgets.

Abstract: Agent memory systems have demonstrated significant potential in long-term dialogue, personalized assistants, and video understanding. However, continuously accumulated memory introduces substantial storage and retrieval costs during inference. To address this issue, we propose \textbf{MemForest}, a general memory compression framework adaptable to various agent memory systems. Specifically, MemForest partitions historical memory into event-centric units by leveraging global semantic similarity and local temporal continuity. For each unit, it constructs a maximum spanning tree, termed an EventTree, and progressively merges redundant memory nodes by selecting high-weight edges, reducing storage overhead. Furthermore, we introduce an anchor-guided propagation retrieval mechanism that retrieves relevant memory nodes from the temporal neighborhoods of key nodes, improving retrieval accuracy. Extensive experiments demonstrate the effectiveness of MemForest. Under the unimodal Mem0 framework, MemForest retains \textbf{97.1%} of the original performance while compressing \textbf{50%} of historical memory across three benchmarks (LoCoMo, LongMemEval, and PersonaMem), achieving a \textbf{1.89x} retrieval speedup. Under the multimodal M3-Agent framework, it preserves \textbf{99.7%} of the original performance with a \textbf{50%} compression ratio across two benchmarks (M3-Bench-robot and M3-Bench-web), achieving a \textbf{2.24x} retrieval speedup. \textcolor{RoyalBlue}{\textit{Our code is available at [https://github.com/Celina-love-sweet/MemForest.}}](https://github.com/Celina-love-sweet/MemForest.}})