Video Reasoning

Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting clip highlights subtle abstract patterns like unwritten rules or hidden labels. This capacity reflects a form of humanity's sixth sense: an intuitive reasoning mechanism that recovers implicit information beyond raw sensory perception. Crucially, this rapid, zero-shot visual intuition underpins everyday navigation and social interaction, making it a vital capability for Multimodal Large Language Models (MLLMs) deployed alongside people. Existing visual benchmarks, however, target either deliberate expert-level analysis in academic and mathematical domains or low-level perception, leaving the intuitive reasoning that people perform largely untested. To bridge this gap, we introduce Humanity's Sixth Sense (HSS), a benchmark for intuitive visual reasoning. HSS spans diverse image and video inputs, organizes items under a structured taxonomy, and pairs each with human-written prompts probing the implicit temporal, spatial, social, and abstract structure that people infer at a glance. Frontier MLLMs fall short of human performance: participants reach 93.1% accuracy, while the strongest model, GPT-6-astra, reaches only 53.6% even at maximum reasoning effort. Despite excelling in many complex tasks that require advanced perception and knowledge, current models still struggle significantly on these visual tasks that are intuitive for humans. We further explore agentic setup that apply dynamic visual manipulation to HSS, which narrows but does not close the gap. HSS establishes intuitive visual reasoning as a measurable axis and directs attention to a capability that scaling on current benchmarks has so far left behind.

Long-video agents can actively gather question-relevant evidence, but they typically leave a central decision implicit: when has the agent seen enough to answer? We propose CASE, a plug-in termination framework that frames this decision as policy-conditioned sequential stopping. At each causal checkpoint, CASE combines an auxiliary multiple-choice assessment of accumulated evidence with the host agent's execution state. From complete native trajectories, we construct a cost-aware target that compares answering now with stopping later along the same search path, accounting jointly for answer correctness and the full cost of continued reasoning. A lightweight Ridge regressor learns this decision gap and produces STOP/CONTINUE decisions. We evaluate three vision-language models with VideoSeek and AVP. On Video-MME, end-to-end accuracy changes by +0.67 percentage points on average while CASE reduces model-token use by 53.63%. The same frozen policies then transfer zero-shot to LongVideoBench and MLVU, with end-to-end accuracy changes of +3.38 and +4.58 points while saving 58.78% and 51.28% of model tokens, respectively. Across all agent-model-benchmark combinations, CASE attains the highest accuracy-efficiency Pareto-frontier coverage among the compared stopping methods (83.3%) at the selected operating points. Online execution preserves this favorable accuracy-efficiency trade-off and additionally reduces measured runtime by 54.1% on average. CASE provides a plug-in termination framework for long-video reasoning agents, enabling them to decide when further evidence acquisition is no longer worthwhile.

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

Reasoning through video generation offers a promising path toward visual intelligence by modeling latent visual states and their dynamics. However, current video generation models often lack explicit guidance on how these states should evolve, leaving generated trajectories prone to physical and structural inconsistencies that undermine reasoning reliability. While the Video Joint-Embedding Predictive Architecture (V-JEPA) provides rich spatiotemporal priors learned through latent prediction, these general priors do not naturally adapt to the logical reasoning capabilities required for complex visual tasks. To bridge this gap, we propose VR-JEPA, a framework that aligns the V-JEPA predictor with task-specific reasoning logic through localized contrastive-state learning and uses its predicted latent trajectories to guide video generation for visual reasoning. Specifically, (i) we pair successful trajectories with generated alternatives under the same input conditions and use discrepancies in their V-JEPA representations to identify informative states and tokens for localized contrastive supervision. (ii) We further equip the V-JEPA predictor with skill-specific experts trained on anchor-task data, allowing the model to adaptively specialize its shared spatiotemporal priors across diverse cognitive domains. Together with skill-specific experts, this contrastive supervision enables VR-JEPA to predict latent trajectories that provide task-specific logical guidance for video generation. Comprehensive experiments on the large-scale VBVR-Pro-Bench dataset demonstrate that VR-JEPA achieves an $11.33\%$ relative improvement over the cutting-edge generation-based reasoning baseline, significantly mitigating physical artifacts and enhancing logical consistency.

Recent advances have enabled unified omni-modal models in understanding audio, vision, and language. However, existing benchmarks, training data, and learning methods largely treat the modalities independently, leaving the capability of audio-visual joint reasoning poorly evaluated and insufficiently elicited. We address this gap with a benchmark, data engine, and learning method. First, we introduce OmniReasoningBench, a benchmark where both audio and visual evidence are indispensable. It comprises 1,150 multiple-choice and open-ended questions across two tasks, reasoning over video and reasoning beyond video. Second, we develop a data engine OmniQA. It automatically constructs evidence-grounded QA pairs that explicitly necessitate audio-visual joint reasoning, together with time-stamped clue chains that guide the annotation of thinking process. Besides our benchmark, this engine produces training data OmniReasoning-SFT-112K and OmniReasoning-RL-19K. Finally, we propose an on-policy self-distillation method Modality-Factored Self-Distillation (MFSD). It evaluates each sampled response under modality-specific clue contexts, disentangling the contributions of individual clues and their cross-modal interactions for token-level credit assignment. With our training data and learning method, our model OmniReasoning-30B-A3B achieves 50.0% on OmniVideoBench and 42.5% on OmniReasoningBench, improving the base model Qwen3-Omni-30B-A3B-Thinking by 12.8 and 9.3 percentage points, respectively. Moreover, it delivers substantial gains on general and long-video benchmarks, including Video-MME-v2. We hope our work offers a solid step for facilitating future research in omni-modal joint reasoning.

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

Embodied systems must make knowledge acquired during one encounter usable in another despite changes in viewpoint, motion, and illumination. Yet aggregate cross-video accuracy conflates failures of local perception with failures to preserve observation identity, establish correspondence, and compose evidence, obscuring whether local video understanding actually transfers. We introduce EgoGears, a complementary single- and multi-video benchmark designed to diagnose this transition. It contains 567 single-video and 1,487 multi-video questions derived from 126 human-collected egocentric recordings covering 39 outdoor routes. Repeated traversals across movement speeds and lighting conditions ground comparisons in shared physical environments; 531 questions require alignment across independent recordings. Single-video questions measure the local visual, spatial, and motion evidence available to a model, while multi-video questions test whether evidence remains bound to the correct observation and can be composed into consistent route relationships. We report 29 single-video and 31 multi-video MLLM configurations across six model families in the main leaderboard. Among the 20 configurations evaluated comparably on both splits, every model performs worse on multi-video questions, with a mean decrease of 22.5 percentage points, and the gap persists when answer format and scoring are held fixed. The gap is not explained simply by additional videos or recording boundaries. The central bottlenecks are observation--evidence binding and ordered route-state tracking. The code and benchmark are publicly available at https://github.com/lei-qi-233/EgoGears.

Reasoning across videos requires aligning events, matching identities, comparing motion, and integrating partial observations. Evaluating these capabilities and testing how to improve them requires both reliable labels and targeted supervision. We introduce SYNCR, a simulator-grounded framework that connects these two needs through shared task generators. Built on Habitat, Kubric, and CLEVRER, SYNCR derives answers from environment state and provides 4,000 evaluation questions and 15,960 training questions over disjoint videos, spanning eight cross-video reasoning tasks. Visual ablations and human evaluation assess dependence on the supplied evidence and answer recoverability. Evaluation of 22 multimodal large language models reveals persistent difficulties in physical comparison and scene integration that increasing model size does not consistently resolve. Supervised fine-tuning raises Qwen3-VL-8B's average SYNCR accuracy from 32.6% to 61.6%, with gains extending to task configurations and video sources absent from training for those tasks. Transfer to real footage is most consistent for temporal ordering: accuracy improves by 9.0-20.5 percentage points on constructed Assembly101 and Panoptic ordering sets across three checkpoints spanning two model families and two model sizes, with additional gains on existing temporal reasoning benchmarks. These results establish SYNCR as a controlled setting for diagnosing cross-video reasoning failures, testing their learnability, and identifying where synthetic supervision transfers.

Streaming video assistance requires models to answer asynchronous questions from an observed prefix under a fixed context budget. Existing approaches model response timing or compress history, but an online state formed before future questions are known can omit visual details before later questions reveal their relevance; the retained state alone cannot recover them. We introduce Watch-Think-Interact (WTI), a closed-loop framework for multi-question streaming video reasoning. WTI maintains compact natural-language memory entries tagged with source-video time ranges; these entries support direct reasoning when sufficient and otherwise anchor selective recall of finer visual evidence. For each question, WTI answers when current context and memory suffice, continues watching when required evidence has not appeared, or recalls a relevant past interval and decides again after incorporating the returned chunks, without replaying the full observed history. To train this behavior, we construct WTI-82K, comprising 82,335 timed questions across 4,812 causally aligned trajectories, and develop Stream-GDPO to optimize complete multi-question streaming rollouts using trajectory-level feedback for response timing, source-video recall, and memory updates. WTI achieves state-of-the-art aggregate performance among the compared open-source streaming baselines, reaching 83.3% on StreamingBench and 73.6% weighted overall accuracy on OVO-Bench.

Video generation models have demonstrated emerging zero-shot capabilities for visual reasoning, perception, and other vision tasks. However, diffusion-based video generation is inherently stochastic, while many downstream vision tasks are deterministic. Motivated by the effectiveness of self-consistency in chain-of-thought reasoning for large language models, we investigate whether self-consistency can similarly improve diffusion-based video reasoning. We first introduce a training-free test-time scaling method that samples multiple video generations and aggregates their predictions through self-consistency. Specifically, we aggregate extracted paths, locations, or masks from multiple rollouts into a consensus prediction. To reduce the inference overhead of multi-rollout generation, we read out predictions early in the denoising trajectory, which preserves consensus quality while reducing denoising steps by more than half. We further propose Rejection Fine-Tuning (RFT) to distill consensus predictions into the video generation model. The resulting model internalizes the benefit of multi-sample consensus and requires only a single generation at inference time, while substantially outperforming the original model. Experiments on three tasks, including maze solving, visual search, and referring segmentation, show that both our self-consistency inference and consensus distillation dramatically improve video-based perception and reasoning, without requiring ground-truth videos or task-specific verification. For visual search, self-consistency raises task accuracy from 48.4% for a single generation to 99.0%. The distilled model retains much of the consensus benefit with a single rollout. For 4-by-4 maze solving, consensus-based training improves the single-generation strict success rate from 72.0% to 84.0% with the same inference latency.

We study whether scaling video generation enables models to reason about hidden information from the past frames, and at what computational cost. Our controlled benchmark requires predicting nine prescribed moves of an initially solved 2x2x2 Rubik's Cube from a fixed view of three faces. Correct predictions require inferring how actions change hidden states, and the simulator provides exact ground truth for evaluation. Models learn plausible cube geometry early, while correct sticker configurations require substantially more training. Although validation MSE follows approximate power-law scaling, lower MSE loss does not reliably indicate downstream reasoning capabilities. Smaller autoregressive models achieve higher state accuracy with limited compute, while larger models reach higher accuracy after more training. At roughly 0.1 PF-days, the 70M-parameter model correctly predicts the visible sticker configuration in 44.6% of post-action frames, compared with 0.3% for the 1B model, which reaches 83.7% at 3.14 PF-days. Symbolic state supervision raises the 20M model's frame accuracy from 31.1% to 67.3% at the same training-data budget, suggesting that learning representations of state changes can complement scaling.

HopChain has shown on still images that multi-hop data synthesis improves vision-language reasoning, because long chain-of-thought reasoning exposes errors that compound across steps, while most data used for reinforcement learning with verifiable rewards (RLVR) rarely demands a chain of visual evidence, so these weaknesses are likely to stay unexposed. We observe the same problem in video, where this framework has not yet been explored. We therefore build Video-HopChain, a dataset of 22,550 multi-hop video questions over 13,378 videos, together with a held-out benchmark of 1,000 questions. Each question chains three to six yes/no questions about moments in one video, and each yields one of two integers depending on its answer. The final answer is the sum of these integers, so an exact match on that sum gives the verifiable reward that RLVR needs. We first train Qwen3-VL-8B with GRPO on a standard video dataset, and a second stage on Video-HopChain then raises the mean over eight video understanding and reasoning benchmarks from 55.4 to 57.9 and improves every one of them. Training on such a dataset, however, exposes a known limitation of GRPO: its learning signal comes from the reward variance within a group, so hard questions whose rollouts are all incorrect and easy questions whose rollouts are all correct both leave the group with no gradient. To recover these groups at the same compute budget, we introduce Confidence-Gated Exploration (CGE). With 8 rollouts per question, CGE samples the first 4 as usual. If these 4 are either all correct or all incorrect, it samples the last 4 with the policy's most confident token masked inside the reasoning span, and removes the masked positions from the loss while all 8 rollouts enter the advantage. With CGE, the mean rises further to 59.3. We release the dataset, the checkpoint, and the data generation and training code.

Real-world video understanding requires integrating visual, audio, textual, and temporal evidence distributed across a video. Yet many pipelines use a fixed video context and single-pass inference, limiting adaptive evidence acquisition when observations are incomplete, ambiguous, or conflicting. We present VLX-VR, an agentic-aware video reasoning model trained within a video reasoning framework defined by a Think--Memory--Observation loop. At each step, VLX-VR determines the needed evidence, invokes read_memory or write_memory, incorporates the returned Observation, and decides whether to continue or produce the task output. We train VLX-VR with multimodal data, including videos and agent trajectories, using reinforcement learning to learn evidence acquisition, memory use, and termination. On MINERVA, VLX-VR achieves state-of-the-art performance among the models included in our comparison, with 78.79% accuracy. Under the original three duration groups, its accuracies are 76.70%, 78.73%, and 80.92%, with a cross-duration accuracy variance of 2.97~$\mathrm{pp}^2$. On correctly answered samples, 96.20% of VLX-VR's reasoning traces are consistent with the MINERVA reference reasoning traces and the evidence described by them, while approximately 75.80% of all evaluated samples satisfy both answer correctness and this evidence-grounded trace criterion. These results show strong performance and broadly stable behavior across durations, while counting, state changes, causal reasoning, and spatial perception remain challenging.

Video understanding demands a convergence of complementary capabilities across perception, temporal understanding, and complex reasoning, which are difficult to jointly optimize within a single model. We introduce Video-MOPD-8B, an open-weight model dedicated to video understanding tasks. To fundamentally enhance its capabilities, we conduct targeted reinforcement learning (RL) optimization across three core domains: video temporal grounding (VTG), general video comprehension, and video STEM reasoning. We then unify their complementary capabilities via Multi-Teacher On-Policy Distillation (MOPD), which consolidates expert knowledge by supervising student-generated trajectories with routed teacher feedback. We further introduce Reliability-Aware Informative Sampling (RAIS), which selects examples with consistently reliable teacher supervision and large teacher-student performance gaps. Together, these components enable Video-MOPD-8B to achieve coordinated and comprehensive performance gains across diverse video understanding tasks. Extensive experiments on comprehensive benchmarks covering general video understanding, temporal grounding, video reasoning, and video STEM tasks demonstrate that Video-MOPD-8B achieves state-of-the-art performance among existing models at a comparable scale. The trained model weights are available at https://huggingface.co/LandH/Video-MOPD-8B.

Despite the remarkable prowess of Vision-Language Models (VLMs) in general multimodal tasks, they remain fundamentally flat'' when reasoning about the physical world. We argue that this spatial bottleneck stems from a profound dimensional mismatch: while VLMs are trained to interpret 2D projections, true spatial reasoning demands the recovery of latent 3D geometry and temporal continuity. To conquer this high-dimensional complexity, we advocate a shift from monolithic learning to a divide and conquer'' paradigm. We present FactoSR, a factorized reinforcement learning framework that explicitly interpret the dimensions collapsed by visual projection. At its core, FactoSR decomposes the monolithic problem of world-consistent reasoning into three orthogonal, geometric sub-objectives: planar correspondence ($XY$), depth consistency ($Z$), and temporal reversibility ($T$). By optimizing these verifiable constraints within a unified policy learning mechanism, we effectively transform an ill-posed projection recovery problem into a series of tangible reasoning steps. Extensive evaluations on multi-view and video benchmarks demonstrate that this elegant decomposition yields substantial gains in 3D and 4D reasoning, achieving a 5.9% boost on VSI-Bench and 4.5% on All-Angles-Bench. Our findings suggest that reinforcing explicit, factorized 4D consistency is a critical step toward evolving VLMs into robust, world-aware reasoners.

We present FoldingAgent, an agentic framework for inferring explicit parametric folding programs directly from origami demonstration videos. Our framework leverages the reasoning power of a pre-trained Vision-Language Model (VLM) equipped with a suite of specialized tools that enable the agent to simulate geometric transitions, verify physical plausibility, retrieve and compare visual content, and evaluate its own predictions. To translate visual content into folding programs, we define a parametric space that consists of the paper's geometry and a set of parametric folding actions. Unlike models that predict static crease patterns, our agent operates sequentially and possesses the ability to re-plan its actions, effectively mitigating the compounding errors inherent in multi-step folding. Our approach takes a step toward closing the gap between human origami knowledge, which is primarily shared through unstructured visual demonstrations, and computational methods, which typically rely on structured, parametric representations such as a crease pattern or an executable parametric plan. We evaluate our approach on PurelandFold, a newly curated benchmark of diverse Pureland origami videos with ground-truth geometry and action labels. Our results demonstrate that by combining VLM reasoning with a set of specialized tools and physical simulation, we can successfully transform unstructured visual demonstrations into executable, physically plausible folding procedures.

Recent advances in Large Multimodal Models (LMMs) have greatly improved video understanding, yet their ability to reason about human-centered social situations remains limited. Existing benchmarks typically rely on videos with a single observed trajectory, making it difficult to determine whether models truly understand social dynamics or merely exploit recurring narrative patterns. We introduce SocialReasonBench, a video multiple-choice QA benchmark for evaluating socially grounded reasoning in scenarios derived from interactive narratives. Built from gameplay videos of Detroit: Become Human, the benchmark leverages branching storylines where player decisions lead to alternative social outcomes that can be checked against the game's own script, flowchart, and recorded branches. We develop a multi-agent curation pipeline that localizes socially meaningful clips, grounds answer labels in game-state signals, and generates theory-guided questions with diagnostic distractors. SocialReasonBench covers seven reasoning dimensions, including intent recognition, emotional empathy, moral dilemma, counterfactual reasoning, and causal antecedent. Experiments on contemporary LMMs show that models perform reasonably well on basic social understanding but struggle with counterfactual and causal reasoning. Further ablation and diagnostic error analyses reveal that models often depend on incomplete modality cues and fall into reasoning traps such as visual shortcuts, highlighting a gap between observable event recognition and deeper reasoning over latent social states.

Large Multimodal Models (LMMs) for video reasoning have long been hindered by the high computational cost of processing vast amounts of visual information. This dilemma motivates the transfer of the reasoning capabilities of large models to smaller, more efficient ones. On-Policy Distillation (OPD) offers a promising solution by matching output-token distributions along student-generated trajectories. However, video reasoning often depends on evidence accumulated across multiple frames. In this context, output-level supervision only captures information expressed through token predictions and does not directly constrain the latent representations formed during reasoning. To address this limitation, we propose Latent-OPD, which augments OPD with trajectory-level latent distillation. Specifically, our method focuses on the position at the end of each trajectory, where hidden states effectively summarize the accumulated visual evidence and reasoning context. Furthermore, we introduce a progressive teacher-lookahead strategy, which aligns middle-to-late student layers with increasingly deeper teacher layers. Experiments on six video reasoning benchmarks show that Latent-OPD consistently outperforms output-only OPD. Notably, the improvements are particularly pronounced in scenarios with limited frames, long videos, or tasks requiring complex evidence aggregation. These results establish Latent-OPD as a highly effective approach to frame-efficient video reasoning.

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

Motion-centric video reasoning is fundamental to interactive applications such as robotic manipulation and autonomous navigation. However, multimodal large language models (MLLMs) typically process videos through sparse uniform sampling to control visual-token and attention costs. This strategy may discard critical transitions between sampled frames, limiting reasoning about object movement, collisions, and causal interactions. To mitigate this issue, we propose Motion-as-Prompt (MaP), a track-guided cross-frame visual prompting framework. MaP recovers dense point trajectories, selects motion-informative frames, and marks the trajectories accumulated between consecutive sampled frames directly onto the visual inputs, making otherwise hidden displacement, direction changes, and interactions observable to frozen MLLMs. Experiments on CLEVRER and Something-Something-v2 show that MaP consistently improves average motion-reasoning accuracy, yielding gains of 4.2% and 8.9% for GPT-5.5, respectively. Notably, these improvements are obtained without degrading non-motion understanding, highlighting the robustness of MaP. These results demonstrate that MaP provides a simple and effective solution for enhancing motion-centric video reasoning without model training or architectural modification. Project page:https://github.com/SunVictor23/MaP.

As generative multimedia evolves from static image synthesis to complex, interleaved visual narratives, a foundational bottleneck has emerged: the judgment crisis. While human perception naturally synthesizes the temporal and logical flow of a story, automated evaluation systems remain largely "blind" to sequential continuity, often failing to distinguish between a coherent narrative and a semantically shuffled or contradictory sequence. This work identifies a critical structural gap in current multimodal evaluation paradigms, arguing that the reliance on Large Vision-Language Models (LVLMs) as judges is fundamentally limited by architectural biases. Our analysis reveals a profound performance dichotomy: while models may appear competent in isolated pointwise scoring, they suffer a catastrophic collapse when required to perform pairwise discrimination of temporal order. We demonstrate that this is not merely a data-scarcity issue but a structural one. Through a series of diagnostic probes, we uncover systematic positional asymmetries, specifically primacy and recency effects, where a model's judgment of a story is significantly influenced by the placement of a frame, often more than by its semantic consistency. These biases, potentially rooted in causal masking and rotary embeddings, suggest that current transformer-based judges are inherently ill-equipped for long-form visual reasoning. By exposing these blind spots, we challenge the multimedia community to move beyond snapshot-centric metrics and instead pioneer Temporally-Aware Evaluation paradigms that treat visual sequences as unified logical structures rather than unordered collections of frames.

Ultra-long egocentric video understanding requires reasoning over temporally sparse evidence distributed across hours or days, challenging current multimodal models with limited context and the grounding of key video segments. While Chain-of-Tool-Thought (CoTT) agent systems enable iterative retrieval and inspection, they suffer from error propagation due to rigid zoom-in strategies that lack recovery mechanisms. In this work, we address these challenges through SCOUT (Self-Checking Chain-Of-Tool-thought), a recovery-aware agentic framework introducing an adaptive policy that evaluates intermediate tool observations and dynamically trades off exploitation (zoom-in) and exploration (region switching), enabling robust multi-hop reasoning over extremely long horizons. However, training such multi-turn tool-using agents remains challenging, as existing RL methods rely on sparse outcome-level rewards and lack supervision over extended decision trajectories, resulting in suboptimal credit assignment for long-horizon reasoning. To address this, we develop UPS-GRPO, an uncertainty-prioritized policy optimization method that concentrates exploration on high-uncertainty post-tool states while preserving sample efficiency. We further introduce a turn-level advantage decomposition that integrates outcome rewards with tool-grounded temporal alignment rewards for improved credit assignment. Experiments show that SCOUT achieves state-of-the-art results on ultra-long egocentric benchmarks, while remaining competitive on shorter-horizon long-video settings.

Sports video analysis is crucial for athletic analytics and broadcasting enhancement. Dense sports video reasoning, however, demands a fine-grained understanding of numerous small-scale, highly interactive, and visually homogeneous entities (e.g., players sharing identical uniforms, the ball) across long temporal contexts. Current Large Multimodal Models (LMMs) inherently struggle with such dense visual complexities. Due to the lack of fine-grained visual details, these models often over-rely on textual priors to guess answers, especially when distinguishing visually similar actions and players. To address this, we propose \textbf{SportsGrounder}, a framework that leverages an open-vocabulary visual expert to aid interleaved grounding specifically for dense sports video reasoning. To achieve precise spatial localization, we extract domain-guided object proposals and introduce an Interleaved Grounding Fusion (IGF) mechanism. The IGF frame-by-frame integrates explicit bounding box coordinates and implicit visual semantics with global grid features. This design preserves strict temporal alignment and prevents sequence length explosion. Furthermore, we design an Action-Aware Supervision (AAS) module that directly regularizes the model's hidden states, forcing the network to learn accurate motion representations rather than relying on language bias. Optimized with Mixed Preference Optimization (MPO) to better distinguish deceptive distractors, our extensive experiments on newly curated dense sports VQA datasets (derived from SoccerNet and FineSports) demonstrate that SportsGrounder significantly improves fine-grained reasoning and achieves state-of-the-art accuracy.

Real-world video reasoning often involves multimodal, multi-source inputs, whereas existing video reasoning tasks typically assume a simplified video-text setting, limiting identity matching and person-centric reasoning. To bridge this gap, we introduce the Identity-conditioned Queries (ICQ) task, in which models are required to jointly associate and interpret an input video and a reference image of a person, and leverage this conditioning to address identity grounding, behavior understanding, and temporal reasoning, among other challenges. Building on ICQ, we present ISYV (I Seek You in Videos), a systematic solution comprising three components: (1) ISYV-Bench, a challenging evaluation benchmark with 1,377 real-world complex videos and 1,377 question-answer pairs, organized into six difficulty levels spanning capabilities from identity recognition to causal reasoning; (2) ISYV-75K, a large-scale training set of 75K high-quality samples constructed via automated annotation, multi-stage verification, and manual review; and (3) ISYV-Framework, containing an ICQ-oriented model and training strategy for learning to exploit informative video shots without additional shot-level annotations. Extensive experiments show that both mainstream closed-source and open-source MLLMs struggle on ISYV-Bench, especially in cross-domain identity matching and long-horizon tracking. ISYV-Model outperforms strong baselines and in some aspects approaches closed-source performance. Overall, ISYV provides a unified task definition, scalable datasets/benchmarks, and modeling insights for person-centric video reasoning.

Spatial intelligence is fundamental to embodied agents, yet existing benchmarks focus on local spatial perception from single or few viewpoints, overlooking global spatial awareness over continuous, long-horizon visual streams. To address this limitation, we introduce the Global-Spatial-Temporal Benchmark (GST-Bench), a VQA benchmark for global spatial intelligence in video understanding, comprising human-verified questions derived from 6,790 minutes of synthetically generated video. It requires models to perform accurate spatial inference from novel viewpoints unseen in the input video and to map egocentric observations onto global top-down images. A comprehensive evaluation of 22 state-of-the-art VLMs exposes a striking gap between models and humans: the strongest zero-shot model attains only 42.68, far below the human score of 79.08. To probe the cause of this gap, we construct GST-Bench-Local and find that models, despite strong local spatial understanding under the same task formulation, still fail to consolidate long-horizon observations into a globally consistent scene representation. We further provide GST-Train, a dataset for global spatial reasoning, as a complementary resource to facilitate future research on this challenge.

Multimodal large language models excel at passive perception but struggle with complex visual cognitive tasks requiring multi-step temporal reasoning. This degradation largely stems from the inherent ambiguity of language-based reasoning, which often fails to accurately articulate continuous visual transformations. To address this, we propose ChronoVision, a multimodal framework designed to align visual logic with latent imagery. During supervised fine-tuning, a Reconstructive Visual Head predicts the latent representation of the final transformed state, while an ROI Attention Locating module focuses the model on key visual evidence via semantic span queries. In post-training, we apply reinforcement learning with an implicit process grounding mechanism, guided by a composite reward function that evaluates outcome correctness, latent process alignment, and unsupervised visual focus. Furthermore, we introduce Vbvr-VQA, a novel dataset that evaluates temporal tracking by reformulating video reasoning into a strict image-ordering task. Experiments demonstrate that ChronoVision achieves state-of-the-art performance on Vbvr-VQA with 74.8% in-domain and 71.6% out-of-domain accuracy, alongside a strong 55.0% accuracy on IntPhys2, a highly challenging cross-domain benchmark.

Timely anticipation of physical hazards is essential for real-world safety, yet existing MLLM evaluations focus on harmful content or general risks, leaving proactive physical hazard prediction underexplored. Sports provide a well-suited testbed: accident causes span diverse injury dimensions and pre-accident spatiotemporal cues draw on reasoning capabilities shared with broader safety domains such as autonomous driving and fall detection. We introduce SPRINT (Sports Proactive Risk INference Testbed), a benchmark of 2,888 real-world sports videos (2,440 accident, 448 safe controls) spanning 14 sports and 3 environmental settings. Accident videos feature fine-grained annotations of early hazard cues, accident timing, and hierarchical causes; safe videos are manually verified as accident-free and serve to diagnose prompt-induced false alarms. Evaluating state-of-the-art MLLMs under diverse prompts and temporal windows reveals a sharp gap between hazard sensitivity and understanding: the best model exceeds 95% in signaling hazards yet falls below 50% in identifying their causes. Diagnostic experiments further show that explicit danger queries trigger severe false alarms even on hazard-free videos. These findings indicate that current MLLMs exhibit only superficial proactive safety, lacking stable, cause-grounded early warning, and underscore the need for reliable proactive safety in dynamic physical environments. Data and code will be open-sourced upon acceptance.

Video question answering requires models to ground language queries in visual evidence and, when necessary, reason over that evidence across time. Existing methods typically rely on long textual chain-of-thought rationales, even though many questions can be answered as soon as the relevant object, action, or frame is localized. We propose Dynamic Latent Reasoning (DyLaR), which first grounds a question in a short block of perception latents (continuous hidden states that encode query-relevant visual evidence), and then adaptively decides whether to append reasoning latents (continuous thoughts that reason over this evidence in latent space) before answering. DyLaR learns this behavior by grounding perception latents in verified visual evidence and distilling verified rationales into reasoning latents, followed by reinforcement learning that further refines when to reason. Across nine video benchmarks and four multimodal language model backbones, DyLaR improves average accuracy over same-backbone baselines while generating fewer than 20 tokens per query. On Qwen3-VL-4B, for example, DyLaR improves average accuracy over Qwen3-VL-4B-Thinking from 54.0 to 58.2 while reducing response length from 1,220.7 to 18.5 tokens per query. Ablations further show that grounded perception latents, rationale-supervised reasoning latents, and adaptive routing each improve accuracy.

Chain-of-thought (CoT) reasoning can improve performance on difficult video questions but often wastes decoding tokens on simple ones. We study whether a video multimodal large language model can adapt its reasoning effort to each question. We propose AdaThinkV, an adaptive framework for video reasoning that learns whether to reason explicitly without offline difficulty labels, manually tuned confidence thresholds, or an external router. During reinforcement learning, AdaThinkV samples matched rollouts in explicit reasoning and direct answering modes for each prompt. ThinkGain estimates the prompt-level utility of explicit reasoning by balancing its accuracy gain against additional response length, providing supervision for both conditional response generation and autonomous mode selection. For difficult prompts, limited rollout exploration can yield groups in which every response is unsuccessful and accuracy rewards show little variation, providing insufficient signal for learning. We therefore introduce Variance Recovery Policy Optimization (VRPO), which retains and progressively expands these groups to recover informative signals from prompts that are difficult yet solvable. At inference, AdaThinkV selects a response mode and generates the response in a single autoregressive sequence. Across a unified suite of video reasoning evaluations, AdaThinkV achieves a mean accuracy of 40.79 with an average of 257.20 output tokens, outperforming the strongest evaluated adaptive baseline by 2.98 points while using 22.7% fewer tokens. Project page: https://trilarflagz.github.io/AdaThinkV/

Existing multimodal long-term memory agents use external memory to overcome the limited context available for long videos. However, most methods emphasize what to store rather than how stored memory should be retrieved. When retrieval becomes inaccurate or repeatedly fails to obtain useful evidence, existing agents lack mechanisms to diagnose failures from previous task trajectories and adapt future search strategies.We introduce Reflective Retrieval Memory (RRM), a reflective memory framework for long-horizon multimodal reasoning. RRM augments an entity-centric multimodal memory graph with reflective experience memory, which distills transferable procedural retrieval knowledge from historical task trajectories. Unlike episodic and semantic memories that preserve factual evidence from the current video, reflective experience memory captures reusable search strategies across tasks. RRM converts retrieved experiences into query-level guidance, while answer generation remains conditioned only on factual evidence newly retrieved from the current video. A lifecycle management mechanism further regulates experience memory through usage frequency, reuse feedback, and temporal decay, thereby reducing redundancy and noise. RRM consistently outperforms previous state-of-the-art approaches on M3-Bench-Robot, M3-Bench-Web, and Video-MME-Long, demonstrating the effectiveness of reflective retrieval memory for long-horizon multimodal reasoning.

Reliable visual safety understanding in real-world scenarios demands more than just object recognition; it requires causal reasoning under epistemic uncertainty. While Large Vision-Language Models (LVLMs) demonstrate impressive semantic alignment on standard benchmarks, they often struggle to distinguish between superficial correlation and genuine forensic logic when grounded in the dynamic, partially observable nature of first-person experiences. Existing evaluations, dominated by third-person surveillance footage and binary classification metrics, fail to expose this cognitive gap. To address this, we introduce EgoSafe-Bench, a benchmark specifically designed to probe forensic reasoning in egocentric safety scenarios. It comprises 12,000 unique evaluation samples, generated by pairing each of the 3,000 video clips with a QA chain governed by our proposed Hierarchical Reasoning Evaluation (HRE) protocol. Unlike standard benchmarks, HRE mandates a rigorous reasoning trajectory from initial feature anchoring to blind-spot deduction and intent inference, thereby enforcing logical consistency and penalizing shortcut-based predictions. Extensive evaluations of state-of-the-art LVLMs (e.g., Qwen3-VL, Gemini, VideoLLaMA 3) reveal a significant perception-reasoning decoupling: models often achieve high descriptive scores but exhibit notable fragility in causal reasoning and logical closure. Our work provides both a challenging dataset and a systematic evaluation framework to foster the development of logically robust video understanding systems.

In the age of foundation models, a model is only as good as its prompt. For this reason, prompt engineering has become an essential technique for improving language model performance. Since video models are currently becoming foundation models for visual tasks (e.g., visual reasoning), we here ask whether they similarly benefit from visual prompt engineering: automatically modifying the task image to improve model performance. For example, for a visual physics reasoning task ("Where does the ball land, after passing a set of obstacles?"), an abstract sketch-like scene can be turned into a photorealistic version with a simple call to an image editing model. We find that visual prompt engineering, or VIPE for short, improves video reasoning performance across tasks. In fact, for video models, visual prompt engineering can be even more effective than classic text-based prompt engineering or test-time scaling. Ultimately, just as text-based prompt engineering systematically improves language model performance, visual prompt engineering can serve as a simple, compute-efficient approach to elicit better visual reasoning performance from video models. Example videos on our project page at https://visual-prompt-engineering.github.io/.

Long-video understanding increasingly relies on large vision-language models and tool-augmented reasoning, but most systems apply the same inference procedure to every example regardless of difficulty. This uniform strategy invokes unnecessary tool-assisted processing for easy questions and provides limited control when difficult questions require fine-grained temporal evidence. We propose CADER (Confidence-Aware Dynamic Evidence Reasoning), a training-free framework for adaptive and reliable long-video reasoning. CADER first performs global reasoning over uniformly sampled frames and estimates answer confidence with a logit-margin signal, allowing high-confidence examples to exit early. For uncertain examples, CADER activates a second-stage tool-augmented loop that combines temporal cropping, lightweight semantic verification, and Relevance-Guided Resampling to progressively localize question-relevant evidence. This design treats tool use as a sample-level decision: a single global pass handles easy cases, while additional reasoning is reserved for examples where uncertainty suggests that more evidence is needed. Experiments on multiple VideoQA benchmarks show that CADER improves long-video reasoning while bypassing Stage~2 for high-confidence samples. Moreover, when applied to a backbone trained only with tool-free chain-of-thought supervision, CADER achieves competitive performance against specialized tool-augmented frameworks, suggesting a practical inference-time route for adaptive long-video reasoning.

Assessing the severity of a traffic accident scenario is important to decide which emergency service to dispatch. Missing an ambulance dispatch on a pedestrian accident is a fatal issue that can lead to death. Recently, Vision-Language Models (VLMs) have been a promising tool for accident reasoning, yet many VLMs are not grounded in real-life accident response protocols, making them not usable in accident severity assessment off-the-shelf. We introduced DispatchRAG, an accident assessor and dispatcher framework grounded in real-life Japanese traffic-accident response protocols, designed to enhance VLMs to generate an appropriate emergency response during an emergency scenario. Utilizing a RAG-based retrieval mechanism to retrieve the most relevant accident protocol and an LLM-powered reasoner to suggest the most proper response. To support evaluation, we introduce Accident Dispatch Dataset, a comprehensive dataset of accident assessment and emergency response according to Japanese accident response protocols adapted from the MM-AU dataset. We validate our framework on the Accident Dispatch Dataset, showing strong performance across various accident scenarios compared to the baseline VLM, pointing toward integration in autonomous vehicles that can automatically report both their own and nearby accidents.

Comprehensive basketball video understanding requires resolving not only what event occurs, but also who is responsible and when the key evidence appears. However, exist- ing methods typically treat spatial perception and semantic recognition as isolated tasks, failing to ground events to individual players or pinpoint their temporal boundaries within complex collective dynamics. To bridge this gap, we introduce BasketEvent, a player- centric basketball event understanding dataset curated from real NBA broadcasts. In BasketEvent, event labels are grounded to the responsible players, and a manually an- notated subset of 1,000 samples with precise event intervals is provided to evaluate tem- poral evidence localization. Based on this data, we propose PlayNet, a player-centric reasoning framework that maps basketball videos to player-level event predictions with temporal evidence. Concretely, PlayNet tracks key entities, associates player identities, and reasons about events by modeling player-player, player-ball, and global court inter- actions, while aggregating sparse temporal evidence via gated pooling. Extensive experi- ments demonstrate that PlayNet significantly outperforms representative video-level and crop-based baselines, proving the superiority of player-centric modeling for fine-grained sports video understanding. Our data, code, and models will be made publicly available.

Cognitive remediation tasks often require patients to perform structured actions involving object manipulation and sequential reasoning. For patients diagnosed with schizophrenia, these tasks are crucial for addressing severe cognitive deficits. However, evaluating the correctness of these physical actions generally relies on manual observation by clinicians, which introduces subjectivity and limits the scalability of therapeutic interventions. In this paper, we propose an automated framework based on Vision-Language Models for action verification in cognitive remediation tasks tailored for schizophrenia rehabilitation. The proposed system relies on a camera-monitored tabletop environment composed of structured miniature scenes including roads, a roundabout, a park, and toy vehicles. Patients receive audio instructions describing goal-oriented spatial actions to perform by manipulating a toy vehicle. These interactive physical activities are specifically designed to stimulate targeted cognitive functions, such as sustained attention, motor coordination, spatial navigation, and cognitive flexibility. To verify the correctness of the performed actions without requiring continuous clinical oversight, the system analyzes the video feed tracking the patient's hand and toy movements. A fine-tuned Vision-Language Model interprets the recorded video sequences and generates semantic descriptions of the observed activities, enabling high-level verification of the executed actions with respect to the initial textual instructions. A dedicated dataset of 4634 tabletop cognitive remediation video scenarios was collected to evaluate the proposed approach. Experimental results demonstrate that our specialized framework effectively bridges low-level physical telemetry with high-level clinical feedback, presenting a scalable and objective solution for advanced cognitive rehabilitation.

Video spatial reasoning is essential for navigation-oriented perception and long-video question answering, where models must infer spatial relations across long horizons under changing viewpoints. However, existing multimodal large language models (MLLMs) remain largely semantic-centric, and often fail to reliably aggregate consistent spatial evidence from redundant video observations, leading to inefficient or unstable reasoning. To address these issues, we propose ConsiSpace, a geometry-consistency-aware framework for geometry-sensitive video spatial reasoning that turns spatial consistency into both an evidence organization principle and an explicit post-SFT learning signal. We build a geometry-consistent memory (GCM) including implicit evidence tokens and explicit geometric cues, and leverage efficient organization strategies to compactly preserve task-related spatial evidence. Furthermore, we utilize unified consistency self-supervised reinforcement learning (UC-SSRL) after supervised fine-tuning to improve cross-view stability, with answer-, metric-, and topology-consistency rewards. Extensive experiments on three spatial-reasoning benchmarks, VSI-Bench, OSI-Bench, and MMSI-Video-Bench, show consistent gains, improving the average score by 12.6 points over the strongest baselines.

Recent advances in world models and video generation have given rise to an emerging reasoning paradigm that leverages video generative models to simulate, predict, and reason about real-world dynamics. We redefine this paradigm as Thinking in Video, where video is not merely an output artifact but a medium for constructing, extending, and verifying causal thought. However, this promise remains unverified: convincing rollouts may reflect memorized appearances rather than causal understanding, while existing metrics separate perceptual fidelity from semantic logic. To evaluate whether video generators support such reasoning, we introduce the Causal-Generative Dual-Judge (CGDJ), auditing World Model Consistency from two perspectives. Explicit Causal Perception tests whether a generator reads a video scenario as a reasoning problem through spatio-temporal flattened visual question answering, while Implicit Generative Perception-Prediction Gap evaluates whether it renders the causal consequence as a consistent future video. Applying CGDJ to representative open- and closed-source generators reveals a clear Perception-Prediction Gap: open-source models produce plausible dynamics despite near-zero explicit causal perception, whereas advanced closed-source systems show stronger but still limited alignment between reasoning and generation. Further analysis exposes audio-visual misalignment, where models verbalize correct causal logic more reliably than they render it, challenging the "world simulator" narrative.

Recent Multimodal Large Language Models (MLLMs) achieve strong performance on single-view video understanding benchmarks. However, sports videos involve dense occlusion, rapid motion, and complex interactions that are difficult to resolve from a single viewpoint. In practice, sports events are recorded from multiple camera angles, providing complementary evidence used by referees. Yet, no existing benchmark evaluates MLLMs on multi-view sports video understanding. To address this gap, we introduce SportMV-Bench, a comprehensive benchmark built from official match recordings, through a dedicated pipeline combining LLM-based generation, MLLM-based verification, and human filtering to ensure quality and consistency. SportMV-Bench containing 787 multi-view video bundles and 2592 question-answer pairs across three categories: Perception-Aware Recognition (PAR), Rule-aware Event Interpretation (REI), and Adjudicative Decision Reasoning(ADR). Our analysis shows that current MLLMs fail to effectively exploit multi-view information, with the bottlenecks lying in fine-grained visual perception and view selection rather than logical reasoning or domain knowledge. We propose SportMV-Agent, an agentic framework that orchestrates an iterative loop of active view selection, perception tool execution, and evidence-grounded reasoning, achieving a significant 14.46% relative improvement over the strongest MLLM baseline.

Automated understanding of complex soccer scenarios from video remains a significant challenge for contemporary vision-language models (VLMs), which suffer from shallow cross-modal alignment and exhibit fundamental limitations in multi-step reasoning and coordinated tool integration. We present TreeSoc, a structured reasoning framework that reformulates soccer video question answering as a hierarchical search problem rather than a single-pass prediction. Specifically, TreeSoc employs a dynamic depth-first search (DFS) mechanism that decomposes complex queries into sequentially ordered sub-tasks, enabling iterative reasoning refinement through explicit intermediate states. This tree-structured decomposition naturally supports adaptive tool routing, wherein domain-specific modules are selectively activated and their outputs incorporated at each reasoning node to produce contextually grounded predictions. On SoccerBench, TreeSoc achieves state-of-the-art performance, with accuracies of 85.2%, 87.4%, and 82.2% on TextQA, ImageQA, and VideoQA, respectively. Additionally, TreeSoc further demonstrates strong cross-domain generalization, attaining 74.16% accuracy on NExT-QA. These results establish structured, tool-augmented tree reasoning as an effective paradigm for robust video understanding. Code is available at: https://github.com/thanhnhan29/TreeSoc.

Video emotion analysis is typically framed as a static classification problem, treating each clip as an independent labeled unit. However, such a formulation overlooks a key psychological fact: emotions change as a result of cumulative reactions to consecutive causal events. To bridge this gap, we introduce Dynamic Affective Reasoning, the first large-scale benchmark for viewer-centric affect transitions and causal reasoning over consecutive video events. DAR contains 15,087 videos and 36,908 event-aligned affective segments annotated with 27 emotion categories. Unlike existing video-based emotion datasets, DAR presents a new viewer-centric perspective on fine-grained emotional expressions and transitions, and provides dense, temporally grounded, and causally explicit reasoning chains. Based on DAR, we formally define three challenging tasks: affective segmentation, fine-grained emotion classification, and affective reasoning. Complementing this benchmark, we propose DAR-R1, a two-stage framework that combines supervised fine-tuning with Group Relative Policy Optimization. Experiments across 10+ MLLMs show that DAR-R1 sets a new state-of-the-art for dynamic affective reasoning, in terms of both emotional localization and affective reasoning. Project page: https://github.com/Zhang-Zhiyan/DAR.

Reasoning has become a core capability for large models, especially when reliable decisions require understanding logical consequences. Recent video generation models offer a reasoning path distinct from previous Chain-of-Thought (CoT): reasoning can unfold through temporally connected frames, known as Chain-of-Frame (CoF) reasoning. However, existing video generators are primarily trained on general video corpora, still lacking diverse supervision and dedicated designs for CoF reasoning. To address this gap, we introduce OpenCoF, a framework comprising the OpenCoF-17K dataset, a reasoning video dataset spanning 11 task families, and Wan-CoF, a fine-tuned video model for studying whether diverse temporal supervision improves CoF behavior. Across four video reasoning benchmarks, Wan-CoF achieves considerable gains over the Wan2.2-I2V-A14B baseline. Building on this, we empirically explore more advanced designs for CoF capabilities, i.e., equipping the model with visual and textual reasoning tokens. This mechanism respectively captures low-level visual cues and high-level semantic priors for spatial and temporal reasoning. Through performance comparisons and attention analysis, we examine how these tokens contribute across model depth, denoising steps, space, and time. Our results suggest that stronger video reasoning requires both broad temporal supervision and explicit mechanisms for organizing intermediate reasoning state. We open-source the dataset, model, and code to facilitate future research on reasoning-oriented video generation.

Video reasoning requires models to identify and verify temporally localized evidence within long video sequences. Recent Video Large Language Models (Video-LLMs) have shown promising reasoning abilities when aligned with reinforcement learning, yet existing approaches typically rely on outcome-based rewards that supervise only the final prediction. Such supervision provides limited guidance on how models should discover the relevant temporal evidence during intermediate reasoning. In this work, we propose TimeThink, a reinforcement learning framework that explicitly guides temporal evidence discovery in Video-LLMs. Our key idea is to treat temporal clue steps as the fundamental optimization primitive of video reasoning, where each reasoning step references a candidate time interval in the video. We introduce a step-wise temporal process reward that provides localized credit assignment for these clues and a joint process--outcome optimization objective that balances reasoning fidelity with task correctness. To enable scalable training, we construct TimeThink-RFT-20K, a dataset with automatically derived temporal evidence segments. Extensive experiments across video reasoning, temporal grounding, and general video understanding benchmarks show that TimeThink consistently improves both temporal localization and reasoning performance, achieving state-of-the-art results among open-source video RL models.

Video reasoning segmentation demands pixel-accurate object tracking across hundreds of frames under complex natural language queries, producing dense spatiotemporal tokens whose quadratic self-attention cost makes long-video processing prohibitive. Existing methods address this through token compression, yet typically operate on encoder features lacking temporal context, constraining selection before content redundancy can be reliably assessed. Informed compression requires contextual awareness, but acquiring that awareness at full resolution incurs the same quadratic cost compression aims to reduce. State-space models resolve this constraint, as their linear recurrence selectively conditions each token on temporal context at $\mathcal{O}(T)$ cost, producing representations where content redundancy becomes assessable. Building on this, Selective SpatioTemporal Aggregation and Compression (STAC) enriches features via decoupled bidirectional spatial and causal temporal scanning, leveraging recurrence-derived redundancy for hierarchical compression with adaptive thresholds optimised with segmentation objective. STAC achieves 85% token reduction and 1.8$\times$ speedup while surpassing compression-free baselines on reasoning segmentation benchmarks in a zero-shot streaming-compatible setting. Code is available here.

Understanding human behavior while interacting with the surrounding world is crucial for many applications of embodied AI. First-person videos are particularly informative for this problem, as they well capture how activities reshape the scene over time. However, existing approaches often rely on implicit visual or language-aligned representations, disregarding structured reasoning over the scene dynamic. We argue that explicit, compositional and editable representations of human-environment interactions can play a crucial role for rich grounded activity understanding. To this end, we introduce SG-Ego, a large scale annotation set extending Ego4D with spatio-temporal scene graphs, where relations triplets are consolidated over time into explicit time-evolving descriptions of the scene state. To reason over this representation, we propose GLEN, a graph-based model that operates over scene graph sequences to both align them with textual actions and model their temporal evolution. In addition, we formulate the activity-driven graph-edit forecasting (A-GEF) problem, a novel task that casts scene dynamics as a sequence of structured transformations conditioned on ongoing actions, enabling explicit reasoning about how scenes change over time. We validate our approach across multiple downstream tasks, spanning retrieval benchmarks as EgoMCQ and EgoCVR, as well as long-horizon reasoning benchmarks as EXPLORE-Bench and the newly introduced A-GEF. GLEN achieves strong results compared to raw video baselines and it excels in reasoning settings, typically addressed only with MLLMs, while enabling controllable and structured predictions of scene dynamics driven by human activities. We believe our results establish spatio-temporal scene graphs, together with models that reason over them, as strong compositional and interpretable representations for video understanding and potentially beyond.

Video reasoning requires Large Multimodal Models (LMMs) to remain grounded in dense evidence, yet existing systems largely adopt "read-once, generate-many" paradigm, in which visual grounding weakens during generation. This phenomenon has been widely observed and is known as Visual Anchoring Decay. To fill this gap, we introduce Latent Video Cache (Latent-VC), a recurrent latent visual cache inserted into the decoder to preserve compact visual memories throughout reasoning. The cache is trained with supervised contrastive cache alignment and vision-grounded GRPO with a latent grounding reward, while maintaining strict train-inference alignment through native decoder hidden states. Built on Qwen3.5-9B, Latent-VC consistently outperforms strong CoT and SFT+GRPO baselines across six video benchmarks, with especially clear gains on grounding-intensive and long-video tasks. In addition, it also achieves higher accuracy with substantially shorter responses, suggesting that latent visual caching improves video reasoning by preserving visual evidence rather than relying on longer textual chains.

Long-video reasoning is fundamentally constrained by how models acquire and utilize visual evidence. Existing tool-augmented video frameworks often interleave temporal grounding and answer reasoning within a single trajectory, causing early semantic hypotheses to bias evidence localization. We term this failure mode premature semantic commitment, where biased grounding retrieves incomplete evidence and incomplete evidence further reinforces incorrect reasoning. To address this issue, we propose EFlow, an evidence-first video reasoning framework built upon Qwen3-VL. EFlow explicitly separates temporal grounding and logical reasoning through CoT for Temporal Grounding and CoT for Reasoning, enabling the model to retrieve relevant evidence before answer inference. In addition, EFlow introduces a confidence-aware reflection mechanism that re-evaluates the full video when retrieved evidence is potentially insufficient. We further construct dedicated trajectory datasets and train EFlow through supervised fine-tuning, reinforcement learning, and reinforcement fine-tuning. Extensive experiments across five video understanding benchmarks demonstrate that EFlow consistently improves long-video reasoning performance.

Video anomaly detection (VAD) with multimodal large language models has shown strong potential, yet most existing methods still depend on large-scale annotations or expert-designed priors, limiting their ability to acquire anomaly knowledge with as little human intervention as possible. To address this, we propose Linguistic Relative Policy Optimization (LRPO), which distills group-relative semantic advantages from multiple reasoning trajectories into a linguistically expressed anomaly experience prior, and adapts the model by injecting this prior into the context to steer its output distribution without any parameter updates. LRPO builds two complementary experience representations: general experience captures transferable anomaly preferences across scenarios, while scenario experience models context-dependent anomaly rules for targeted refinement. To further improve the learned experience, we introduce an anomaly alignment reward that guides trajectory optimization to match human risk preferences and reinforce temporally grounded reasoning. Extensive experiments on XD-Violence, UCF-Crime, and UBnormal demonstrate that LRPO significantly outperforms existing state-of-the-art methods under tuning-free settings.

Despite the rapid advance of Multimodal Large Language Models (MLLMs) in high-level video understanding, a fundamental bottleneck remains: these models collapse complex human motion into coarse semantic labels. Existing benchmarks mostly focus on scene-centric events or local joint articulations, failing to probe global human motion in space over time (trajectory and orientation changes). We introduce HumanMoveVQA, the first comprehensive benchmark designed to evaluate global trajectory and orientation reasoning from an exocentric perspective. Our benchmark utilizes a first-frame anchored world coordinate system, preserving translation and rotation relative to a fixed starting point. We propose a scalable, multi-stage pipeline that lifts 2D video observations into world-consistent 3D motion tracks to generate over 10K structured question-answer pairs across seven reasoning categories, including motion aggregation, sequential ordering, and trajectory-level inference. Our extensive evaluation reveals a critical capability gap in state-of-the-art proprietary models on deep human motion understanding. However, we demonstrate that this is a learnable problem; by fine-tuning an open-source baseline with our targeted, world-consistent supervision, we achieve a significant improvement. HumanMoveVQA establishes a rigorous geometric foundation for developing next-generation, movement-aware video understanding models.

Current multimodal reflection mechanisms for long video understanding predominantly rely on closed-loop self-reflection within internal parameters. Lacking objective external evidence, models are frequently trapped in blind confidence and often fail to correct errors. Furthermore, applying reinforcement learning to multi-stage reflection pipelines introduces severe policy coupling, which is exacerbated by a critical scarcity of dedicated training data. To address these limitations, this work proposes Reflect-R1, the first Evidence-Driven self-correction framework for long video understanding. The framework constructs a three-stage pipeline consisting of intuition, verification, and arbitration. By dynamically retrieving objective visual evidence to verify initial intuitions and autonomously executing multiple temporal searches to resolve conflicts, it completely breaks the hallucination loop. To overcome policy coupling, we design a stage-decoupled reinforcement learning algorithm named SD-GRPO that independently computes advantage functions across different reasoning stages. Concurrently, we construct a dataset of 120K samples to bridge the training data gap. Extensive experiments on benchmarks such as VideoMME and LongVideoBench demonstrate that Reflect-R1 achieves state-of-the-art performance. Our method significantly improves the genuine rectification rate and enables authentic self-correction strictly grounded in objective evidence.

Recent interest in multimodal large language models (MLLMs) raises a central question: can they reason over dynamic visual evidence rather than merely recognize objects or events in individual frames? This ability, which we refer to as video temporal-logical reasoning, requires models to maintain, update, and compose evidence as visual states evolve across frames. Existing video benchmarks often conflate this capability with scene complexity, static recognition, or uncontrolled temporal variation. To isolate this capability, we introduce Video-MME-Logical, a controlled benchmark organized around five temporal-logical operations: state tracking, sequential counting, temporal ordering, dynamic spatiality, and structural composition. The benchmark contains 25 fine-grained task categories generated with controlled object states, transitions, temporal dependencies, and logical compositions. It enables difficulty-controlled final-answer evaluation by varying temporal horizon and reasoning complexity, and supports intermediate-state diagnostics by verifying whether models recover the required logical reasoning trace before producing the final answer. Experiments with state-of-the-art MLLMs reveal a substantial human-model gap, especially as temporal-logical complexity increases. Supervised fine-tuning on up to 500K generated samples improves performance but remains insufficient to close the reasoning gap, positioning Video-MME-Logical as a scalable testbed for analyzing and improving temporal-logical reasoning in MLLMs.

Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier video reasoning models can suffer 15-30%p accuracy drops on real-world embodied benchmarks, while remaining unaware that their visual evidence has been degraded. To address this challenge, we propose Robust-TO, an agentic video understanding framework that explicitly integrates per-frame trustworthiness into every stage of reasoning. Robust-TO organizes heterogeneous visual perception tools under a unified evidence interface. Each tool receives a sub-query derived from the original question and a set of trustworthy frames selected by the reliability-relevance score. It returns evidence in a shared format: a concrete prediction (e.g., a bounding box, motion trajectory, recognized text, or action label), temporal grounding, and a calibrated reliability score. During reasoning, these calibrated scores guide evidence weighting in a three-tier synthesis process (high/medium/low) and define a confidence-cost GRPO reward that jointly optimizes correctness, evidence reliability, and efficiency. On two video reasoning benchmarks spanning eight tasks, Robust-TO achieves 56.4% average accuracy on clean inputs, surpassing the strongest open-source baseline by 10.6%p and outperforming Gemini-2.5-Pro (46.2%). Under five realistic corruption types, Robust-TO maintains 54.3% average accuracy, 5.8%p above the strongest open-source baseline, while exhibiting the smallest clean-to-corrupted accuracy drop among all compared methods.

A robot working alongside people must reason about what they have done, in what order, and with what intent. Video carries the spatial layouts, object histories, and gestures that language leaves underspecified, yet today's manipulation benchmarks pair an instruction with a single current image, offering no way to evaluate reasoning over observed human behavior. We introduce WatchAct, a benchmark for robot manipulation grounded in observed human behavior. Each instance pairs a real-world human-action video and a language instruction with an aligned simulator scene and an executable LIBERO task, enabling scalable and reproducible evaluation. WatchAct comprises 3,000 long-horizon instances across 14 tasks in four capability domains drawn from the cognitive demands of watching another agent: parsing events (Event Grounding), recovering procedural structure (Procedural Reasoning), inferring unstated intent (Implicit Intent Inference), and tracking how the scene was changed (Episodic Reasoning). We further propose a disentangled evaluation protocol that separately measures (i)~video-to-plan reasoning by vision-language models, (ii)~policy execution under oracle plans, and (iii)~full task completion by integrated planner--policy pipelines. In both simulation and on a Franka Research 3 robot, current systems remain far from solving WatchAct. The best pipeline, Gemini-3.1-Pro with $π_{0.5}$, reaches only 16.3% Success Rate (SR) in simulation and 14.0% on the real robot. Gemini-3.1-Pro attains just 36.8% Plan SR (vs. 97.1% for humans), while $π_{0.5}$ reaches only 21.5% Task SR under oracle plans and drops to 10.6% on out-of-domain scenarios. Dataset and code are available at https://baiqi-li.github.io/watchact_page/.

Existing Referring Video Object Segmentation tasks focus on referring expressions describing events, actions or appearances of relevant objects within the observed frames, lacking evaluation in scenarios that require pre-decisive spatio-temporal reasoning, thereby limiting their applicability. To address this, we propose Foresight Expression Video Object Segmentation, a task that queries future events in upcoming video segments and requires masks of the objects in the observed frames as visual answers. For example, in ego-centric scenes, the question "What tool will be used?" demands reasoning over spatio-temporal cues to predict the masks of the next tool to be used, which helps with the understanding of future actions and decisions. To support this task, we introduce FeVOS, a dataset with 968 video clips, 14,525 foresight expressions, and 2,904 chain-of-thought annotations to provide explicit and interpretable reasoning steps. We further develop FeVOS-R1, an MLLM-based model trained on our dataset via a two-stage pipeline of supervised fine-tuning and reinforcement learning. FeVOS-R1 not only achieves state-of-the-art performance on FeVOS, but also demonstrates strong generalization to existing RVOS benchmarks. We hope this work can inspire more research on predictive reasoning in video perception.

Video MLLMs often struggle with fine-grained spatio-temporal reasoning, sometimes generating correct answers based on irrelevant frames or objects. Although outputting spatio-temporal evidence during reasoning is a promising direction, existing RL frameworks typically rely on geometry-only (IoU) rewards, which can be sensitive to boundary perturbations and overlook semantic alignment. To address this, we propose Semantic Evidence Reward (SER), which reformulates spatio-temporal evidence grounding as a constrained verification task. Instead of computing pixel-level overlap, SER uses a referee VLM as a local checker to evaluate model-generated evidence claims across two dimensions: relevance and localization quality, combined with a temporal penalty. This design reduces the reliance on dense box annotations and enables training directly on standard video QA data. On the V-STAR benchmark, SER achieves 49.6% mLGM, improving by 3.0 points over the strong evidence-grounded baseline Open-o3-Video, demonstrating its potential in enhancing both answer accuracy and evidence grounding.

Recent advancements in chain-of-thought (CoT) reasoning have shown promise in enhancing video understanding and reasoning capabilities of multimodal large language models (MLLMs). However, existing CoT-based MLLMs require labor-intensive CoT annotations and incur substantial training and inference overhead. While visual latent reasoning has emerged as a more efficient alternative, existing methods primarily focus on image tasks and heavily rely on additional supervision signals for visual latent generation (e.g., CoT traces, auxiliary images, or fine-grained annotations), limiting their scalability and transferability to video tasks. To bridge this gap, we introduce VideoLatent, a novel MLLM equipped with a latent injection module tailored for video understanding and reasoning. Specifically, VideoLatent learns to perform visual latent reasoning using a new latent self-forcing training paradigm, which comprises latent alignment and latent diversity objectives, and relies solely on standard video-question-answer triplets. Extensive experiments across 14 benchmarks demonstrate that our model consistently outperforms existing standard and latent MLLMs on general video understanding and complex video reasoning. Compared with Video-R1, our VideoLatent achieves superior computational efficiency, reducing training/inference overhead by $\sim$6$\times$/$\sim$68$\times$. Moreover, experiments demonstrate that our method has strong generalizability to different MLLM backbones and different model scales.

Understanding long videos requires fine-grained perception and multi-step, higher-order reasoning over complex, long-range spatio-temporal dynamics. Vision-language models (VLMs) encode video frames into visual tokens and attempt to perform both perception and multi-step planning latently, within a single forward pass. This coupled formulation, however, is bottlenecked by the LLM's limited capacity to discover and execute multi-step strategies in its latent representations. To address this bottleneck, we propose Hierarchical Programmatic Probing (HPP), a framework that decouples semantic perception from higher-order temporal reasoning by reformulating long video understanding as iterative, programmatic exploration of a hierarchically segmented video. Specifically, a coding-capable LLM plans and executes a multi-step strategy in an interactive coding environment, probing the video for information and invoking a VLM for localized perception on demand. To make probing tractable over long videos, we introduce three components: information-density-aware hierarchical segmentation, late-interaction semantic retrieval, and structured probing functions for coarse-to-fine temporal localization. We validate HPP on LongVideoBench, which requires both fine-grained perception and long-range relational reasoning, and show that decoupling the two via iterative programmatic probing yields substantial gains. Further results on EgoSchema, VideoMME, and MLVU demonstrate the effectiveness of our approach across diverse long-video benchmarks.

In multimodal video reasoning, reinforcement learning-based methods typically rely on simplistic and inflexible reasoning-length control strategies that fail to adapt to the model's evolving competence. This mismatch may suppress necessary exploration at early stages, while encouraging redundant reasoning and inefficient decoding once the model becomes more competent. In this paper, we propose CARE, a competence-aware reward shaping framework for adaptive reasoning length optimization in multimodal reasoning. Specifically, CARE maintains a smoothed competence estimate via an exponential moving average of pass rates, and uses it to route training into progressive stages that shift the reward preference from exploration-oriented long-form reasoning to efficiency-oriented concise reasoning. To avoid conflating verbosity with intrinsic task complexity, CARE further normalizes reasoning effort with batch-level statistics, and introduces a posterior amplifier to strengthen reward signals for unexpectedly strong performance on historically difficult samples. The proposed mechanism is seamlessly integrated into the GRPO training pipeline and incurs no additional inference-time overhead. Extensive experiments on multiple video reasoning and general video understanding benchmarks demonstrate that CARE consistently improves reasoning accuracy, stabilizes reinforcement learning, and significantly enhances token efficiency. Moreover, CARE exhibits a characteristic inverted-U trajectory of reasoning length during training, and yields shorter yet more informative reasoning traces at convergence, indicating effective adaptive allocation of reasoning budget. We provide the source code for our proposed CARE framework and experiments at https://github.com/1Pansy/Video-CARE.

Reinforcement learning has improved the reasoning ability of large language models, but applying outcome-only rewards to video multimodal large language models (Video-MLLMs) provides limited guidance on which visual evidence should support the answer. Inspired by multisensory integration, where consistent cues can enhance the salience and reliability of perceptual estimates, we introduce Consensus Frame GRPO (CF-GRPO), a temporal-annotation-free process-level reward framework for evidence-aware video reasoning. CF-GRPO constructs a consensus frame prior from intrinsic video cues, including temporal coverage, scene-transition cues, and query-conditioned visual relevance. It then computes a model-side frame-use score from visual and response representations and optimizes their agreement through the Consensus Frame Reward (CFR). With salience-aware sparse aggregation and distribution sharpening, CFR provides a high-contrast reward signal without requiring human temporal annotations. Experiments show that VideoCFR achieves competitive performance across complex video reasoning benchmarks and improves several metrics over representative Video-MLLM and RL baselines, while the consensus prior provides an interpretable view of the evidence frames emphasized during training. The implementation is available at https://github.com/1Pansy/VideoCFR.

Text-to-video retrieval in operating rooms (OR) is an enabling technology for OR safety, as it allows stakeholders to retrieve and inspect recordings of specific events. However, because the most safety-critical events may not follow the common structure, to unlock its full potential text-to-video retrieval must be able to handle implicit queries that require reasoning to identify the right video (e.g., the step right before clipping). However, existing methods rely on global embeddings that cannot reason over such queries. We propose OR3, a text-to-video retrieval method that converts clips into action-driven digital twins (ActDTs), grouping concurrent subject-action-object triplets under non-overlapping temporal intervals. Moreover, rather than cross-modal matching through paired encoders, OR3 performs imagination-based retrieval where an LLM generates hypothetical ActDTs from queries. This enables intra-modal matching via a single encoder trained with ActDT-tailored hard negatives. Finally, evidence-grounded refinement revises imagined ActDTs based on discrepancies with top candidates to capture procedure-specific patterns. We construct a benchmark from MM-OR with 276 implicit queries across four reasoning categories over 386 clips from robotic knee procedures. OR3 achieves 57.6 R@1 and 77.3 R@5, outperforming the strongest baseline. These results demonstrate that OR3 enables fine-grained discrimination between visually similar OR video clips through temporal action reasoning.