While a general embodied agent must function as a unified system, current methods are built on isolated models for understanding, world modeling, and control. This fragmentation prevents unifying multimodal generative capabilities and hinders learning from large-scale, heterogeneous data. In this paper, we propose Motus, a unified latent action world model that leverages existing general pretrained models and rich, sharable motion information. Motus introduces a Mixture-of-Transformer (MoT) architecture to integrate three experts (ie, understanding, video generation, and action) and adopts a UniDiffuser-style scheduler to enable flexible switching between different modeling modes (ie, world models, vision-language-action models, inverse dynamics models, video generation models, and video-action joint prediction). Motus further leverages the optical flow to learn latent actions and adopts a recipe with three-phase training pipeline and six-layer data pyramid, thereby extracting pixel-level "delta action" and enabling large-scale action pretraining. Experiments show that Motus achieves superior performance against state-of-the-art methods in both simulation (a +15% improvement over X-VLA and a +45% improvement over Pi-0.5) and real-world scenarios(improved by +11 48%), demonstrating unified modeling of all functionalities and priors significantly benefits downstream robotic tasks.
论文检索
输入标题、作者或关键词，从 2,006 篇学术成果中精准定位
We revisit video hallucination in multimodal large language models (Video-MLLMs) from a semantic aggregation perspective. While prior work attributes hallucinations to language priors, missing frames, or visual encoder biases, these explanations overlook errors arising during the aggregation of correct frame-level semantics into event-level interpretations. We term this phenomenon Semantic Aggregation Hallucination (SAH), which becomes increasingly prevalent in complex, multi-event video understanding tasks with rich temporal dependencies. To systematically study SAH, we introduce ELV-Halluc, the first benchmark designed for fine-grained evaluation of semantic aggregation errors. Our experiments reveal that SAH correlates with both semantic complexity and rapid semantic transitions. We further propose mitigation strategies: improved positional encoding preserves temporal structure, and reinforcement learning such as DPO enhances the model's ability to distinguish semantics within and across events. Using a curated 8K adversarial video-text pair dataset, our approach achieves consistent gains across benchmarks, including a 27.7% reduction in SAH rate on ELV-Halluc and Video-MME. Data and code are available at https://github.com/hlsv02/ELV-Halluc.
Mimic Human Cognition, Master Multi-Image Reasoning: A Meta-Action Framework for Enhanced Visual Understanding
PDF ↗While Multimodal Large Language Models (MLLMs) excel at single-image understanding, they exhibit significantly degraded performance in multi-image reasoning scenarios. Multi-image reasoning presents fundamental challenges including complex inter-relationships between images and scattered critical information across image sets. Inspired by human cognitive processes, we propose a Cognition-Inspired Meta-Action Framework (CINEMA), which decomposes multi-image reasoning into five structured meta-actions: Global, Focus, Hint, Think, and Answer, explicitly modeling the sequential cognitive steps humans naturally employ. For cold-start training, we introduce a Retrieval-Based Tree Sampling strategy that generates high-quality meta-action trajectories to bootstrap the model with reasoning patterns. During reinforcement learning, we adopt a two-stage paradigm: an exploration phase with Diversity-Preserving Strategy to avoid entropy collapse, followed by an annealed exploitation phase with DAPO to gradually strengthen exploitation. To train our model, we construct a dataset of 56k cold-start and 58k reinforcement learning instances spanning multi-image, multi-frame, and single-image tasks. We conduct extensive evaluations on multi-image reasoning benchmarks, video understanding benchmarks, and single-image benchmarks, achieving competitive state-of-the-art performance on several key benchmarks. Our model surpasses GPT-4o on the MUIR and MVMath benchmarks and notably outperforms specialized video reasoning models on video understanding benchmarks, demonstrating the effectiveness and generalizability of our human cognition-inspired reasoning framework.
Hand-object interaction (HOI) involves dynamics where human manipulations produce spatio-temporal effects on objects. However, existing semantic HOI benchmarks focus on either manipulation or effects at a coarse level, lacking fine-grained spatio-temporal reasoning to capture HOI dynamics. We introduce HanDyVQA, a fine-grained video QA benchmark that comprehensively covers both the manipulation and effect aspects of HOI. HanDyVQA comprises six complementary question types (Action, Process, Objects, Location, State Change, and Object Parts), totaling 11.1K multiple-choice QA pairs. Collected QA pairs require recognizing manipulation styles, hand/object motions, and part-level state changes. HanDyVQA also includes 10.3K segmentation masks for Objects and Object Parts, enabling the evaluation of object/part-level reasoning in video object segmentation. We evaluated recent video foundation models on our benchmark and found that even the best, Gemini-2.5-Pro, achieved only 73% accuracy, well below human performance (97%). Further analysis shows the remaining challenges in spatial relationships, motion, and part-level geometric understanding. We also found that incorporating explicit HOI cues into visual features improves performance, providing insights for future HOI-aware models.
Understanding social interaction in video requires reasoning over a dynamic interplay of verbal and non-verbal cues: who is speaking, to whom, and with what gaze or gestures.While Multimodal Large Language Models (MLLMs) are natural candidates, simply adding visual inputs yields surprisingly inconsistent gains on social tasks. Our quantitative analysis of cross-modal attention inside state-of-the-art MLLMs reveals a core failure mode: in multi-speaker scenes, visual and textual tokens lack speaker-consistent alignment, exhibiting substantially weaker cross-modal attention than in object-centric images.To address this, we propose a multimodal multi-speaker attention alignment method that can be integrated into existing MLLMs. First, we introduce dynamic cross-modal head selection to identify attention heads most responsible for grounding. Then, an adaptive social-aware attention bias, computed from existing attention patterns and speaker locations, is injected into the attention mechanism. This bias reinforces alignment between a speaker's visual representation and their utterances without introducing trainable parameters or architectural changes.We integrate our method into three distinct MLLMs (LLaVA-NeXT-Video, Qwen2.5-VL, and InternVL3) and evaluate on three benchmarks (TVQA+, MMSI, OnlineMMSI). Across four social tasks, results demonstrate that our approach improves the ability of MLLMs and achieves state-of-the-art results.Attention visualizations confirm our method successfully focuses the model on speaker-relevant regions, enabling more robust multi-party social reasoning.
Wide-baseline matching (WBM) requires integrating geometric understanding, viewpoint changes, fine-grained perception, and occlusion reasoning, making it a challenging testbed for spatial reasoning in multimodal large language models (MLLMs) deployed in physical environments. However, current MLLMs lack systematic evaluation and training frameworks for these capabilities. We introduce ReasonMatch-Bench, a benchmark stratified by viewpoint displacement and matching granularity across indoor, outdoor, and object-centric scenarios, and show that current MLLMs still struggle with fine-grained wide-baseline correspondence: on a difficult 90-sample subset, human annotators achieve 84.0 F1, while the best existing baseline reaches 37.2. To bridge this gap, we build a scalable data-generation pipeline that automatically extracts wide-baseline view pairs from large-scale video-3D corpora, including RGB-D videos and SfM reconstructions, yielding diverse and verifiable supervision. We further propose Dynamic Correspondence Reinforcement Learning (DCRL), which combines Image-Level Viewpoint Progression and Point-Level Correspondence Curriculum to improve WBM training through verifiable rewards without explicit CoT supervision. Extensive experiments show that DCRL substantially improves ReasonMatch-Bench and transfers to related spatial benchmarks, while maintaining general visual understanding performance with modest gains on several benchmarks.
From Where Things Are to What They Are For: Benchmarking Spatial-Functional Intelligence in Multimodal LLMs
PDF ↗Human-level agentic intelligence extends beyond low-level geometric perception, evolving from recognizing where things are to understanding what they are for. While existing benchmarks effectively evaluate the geometric perception capabilities of multimodal large language models (MLLMs), they fall short of probing the higher-order cognitive abilities required for grounded intelligence. To address this gap, we introduce the Spatial-Functional Intelligence Benchmark (SFI-Bench), a video-based benchmark with over 1,500 expert-annotated questions derived from diverse egocentric indoor video scans. SFI-Bench systematically evaluates two complementary dimensions of advanced reasoning: (1) Structured Spatial Reasoning, which requires understanding complex layouts and forming coherent spatial representations, and (2) Functional Reasoning, which involves inferring object affordances and their context-dependent utility. The benchmark includes tasks such as conditional counting, multi-hop relational reasoning, functional pairing, and knowledge-grounded troubleshooting, directly challenging models to integrate perception, memory, and inference. Our experiments reveal that current MLLMs consistently struggle to combine spatial memory with functional reasoning and external knowledge, highlighting a critical bottleneck in achieving grounded intelligence. SFI-Bench therefore provides a diagnostic tool for measuring progress toward more cognitively capable and truly grounded multimodal agents.
Scaling the Long Video Understanding of Multimodal Large Language Models via Visual Memory Mechanism
PDF ↗Long video understanding is a key challenge that plagues the advancement of Multimodal Large language Models (MLLMs). In this paper, we study this problem from the perspective of visual memory mechanism, and proposed a novel and training-free approach, termed Flexible Memory (FlexMem). In principle, FlexMem aims to mimic human behavior of video watching, i.e., continually watching video content and recalling the most relevant memory fragments to answer the question. In this way, FlexMem can help MLLMs achieve video understanding of infinite lengths, unlike previous methods that process all video information at once and have input upper-limit. Concretely, FlexMem first consider the visual KV caches as the memory sources, and realize the effective memory transfer and writing via a dual-pathway compression design. Afterwards, FlexMem also explores different memory reading strategies for the diverse video understanding tasks, including the popular streaming one. To validate FlexMem, we apply it to two popular video-MLLMs, and conduct extensive experiments on five long video and one streaming video task. The experimental results show that on a single 3090 GPU, our FlexMem can achieve obvious improvements than existing efficient video understanding methods and process more than 1k frames, which also helps the base MLLMs achieve comparable or even better performance than SOTA MLLMs on some benchmarks, e.g. , GPT-4o and Gemini-1.5 Pro. Our code is released at: https://github.com/city1517/FlexMem.
VideoRealBench: A Chain-of-Thought Realism Evaluation Benchmark for Generated Human-Centric Videos
PDF ↗With the great advancement of video generation models, a growing number of content creators and researchers are leveraging these technologies to produce large volumes of human-centric videos for content creation and customized data generation for specific tasks. Although existing video generation models are capable of producing videos with high visual quality, their inadequate understanding of video realism results in generating unrealistic videos. While various evaluators have emerged to assess the quality of generated videos, they are trained from low-quality generated videos and data annotations, leading to misaligned ratings with human preferences. They also lack interpretability due to the absence of chain-of-thought reasoning. To address these issues, we propose VideoRealBench, a comprehensive benchmark for evaluating the realism of generated human-centric videos. We leverage a rating scale designed from human preferences to score videos and provide three-step rationales, thereby creating a finely-annotated dataset VideoRealDataset and proposing an evaluator VideoRealEval capable of providing reliable scores along with detailed rationales. VideoRealEval achieves a Pearson's linear correlation coefficient (PLCC) of 57.07% and a Spearman's rank correlation coefficient (SROCC) of 56.78% on VideoRealDataset, demonstrating closer alignment with human preferences than existing evaluators. Our code and dataset are available at https://github.com/MCG-NJU/VideoRealBench.
The collection and detection of video anomaly data has long been a challenging problem due to its rare occurrence and spatio-temporal scarcity. Existing video anomaly detection (VAD) methods under perform in open-world scenarios. Key contributing factors include limited dataset diversity, and inadequate understanding of context-dependent anomalous semantics. To address these issues, i) we propose LAVIDA, an end-to-end zero-shot video anomaly detection framework. ii) LAVIDA employs an Anomaly Exposure Sampler that transforms segmented objects into pseudo-anomalies to enhance model adaptability to unseen anomaly categories. It further integrates a Multimodal Large Language Model (MLLM) to bolster semantic comprehension capabilities. Additionally, iii) we design a token compression approach based on reverse attention to handle the spatio-temporal scarcity of anomalous patterns and decrease computational cost. The training process is conducted solely on pseudo anomalies without any VAD data. Evaluations across four benchmark VAD datasets demonstrate that LAVIDA achieves SOTA performance in both frame-level and pixel-level anomaly detection under the zero-shot setting. Our code is available in https://github.com/VitaminCreed/LAVIDA.
Text-guided human body animation has advanced rapidly, yet facial animation lags due to the scarcity of well-annotated, text-paired facial corpora. To close this gap, we leverage foundation generative models to synthesize a large, balanced corpus of facial behavior. We design prompts suite covering emotions and head motions, generate about 80 hours of facial videos with multiple generators, and fit per-frame 3D facial parameters, yielding large-scale (prompt and parameter) pairs for training. Building on this dataset, we probe language models for bidirectional competence over facial motion via two complementary tasks: (1) Motion2Language: given a sequence of 3D facial parameters, the model produces natural-language descriptions capturing content, style, and dynamics; and (2) Language2Motion: given a prompt, the model synthesizes the corresponding sequence of 3D facial parameters via quantized motion tokens for downstream animation. Extensive experiments show that in this setting language models can both interpret and synthesize facial motion with strong generalization. To best of our knowledge, this is the first work to cast facial-parameter modeling as a language problem, establishing a unified path for text-conditioned facial animation and motion understanding.
This paper does not introduce a novel method but instead establishes a straightforward, incremental, yet essential baseline for video temporal grounding (VTG), a core capability in video understanding. While multimodal large language models (MLLMs) excel at various video understanding tasks, the recipes for optimizing them for VTG remain under-explored. In this paper, we present TimeLens, a systematic investigation into building MLLMs with strong VTG ability, along two primary dimensions: data quality and algorithmic design. We first expose critical quality issues in existing VTG benchmarks and introduce TimeLens-Bench, comprising meticulously re-annotated versions of three popular benchmarks with strict quality criteria. Our analysis reveals dramatic model re-rankings compared to legacy benchmarks, confirming the unreliability of prior evaluation standards. We also address noisy training data through an automated re-annotation pipeline, yielding TimeLens-100K, a large-scale, high-quality training dataset. Building on our data foundation, we conduct in-depth explorations of algorithmic design principles, yielding a series of meaningful insights and effective yet efficient practices. These include interleaved textual encoding for time representation, a thinking-free reinforcement learning with verifiable rewards (RLVR) approach as the training paradigm, and carefully designed recipes for RLVR training. These efforts culminate in TimeLens models, a family of MLLMs with state-of-the-art VTG performance among open-source models and even surpass proprietary models such as GPT-5 and Gemini-2.5-Flash. All codes, data, and models will be released to facilitate future research.
Recent advancements in video models have shown tremendous progress, particularly in long video understanding. However, current benchmarks predominantly feature western-centric data and English as the dominant language, introducing significant biases in evaluation. To address this, we introduce CURVE, a challenging benchmark for multicultural and multilingual video reasoning. CURVE comprises high-quality, entirely human-generated annotations from diverse, region-specific cultural videos across 18 global locales. Unlike prior work that relies on automatic translations, CURVE provides complex questions, answers, and multi-step reasoning steps, all crafted in native languages. Making progress on CURVE requires a deeply situated understanding of visual cultural context. Furthermore, we leverage CURVE's reasoning traces to construct evidence-based graphs and propose a novel iterative strategy using these graphs to identify fine-grained errors in reasoning. Our evaluations reveal that SoTA Video-LLMs struggle significantly, performing substantially below human-level accuracy, with errors primarily stemming from the visual perception of cultural elements. We will release CURVE to foster the development of more equitable and capable multimodal foundation models.
FAVE: A Structured Benchmark for Fine-Grained Audio-Visual Temporal Evaluation in Multimodal LLMs
PDF ↗Audio-visual large language models (AVLLMs) have made significant strides in understanding visual and auditory content. However, their ability to capture fine-grained temporal relationships between audio and visual streams remains insufficiently evaluated. To address this, we introduce FAVE (Fine-grained Audio-Visual Temporal Evaluation), a comprehensive benchmark targeting three core dimensions of temporal perception: cross-modal temporal alignment (FAVE-align), event temporal relationship (FAVE-low), and detailed moment captioning (FAVE-high). To construct FAVE, we propose a scalable annotation pipeline that integrates shot boundary detection, automated captioning, and GPT-assisted refinement to produce temporally grounded, high-quality data. Extensive experiments on twelve state-of-the-art multimodal LLMs, both open-source and closed-source, reveal key limitations in multimodal integration, temporal relationship and timestamp localization, especially for joint audio-visual tasks. These findings highlight the need for better temporal modeling to improve AVLLMs' understanding of real-world video content. FAVE serves as a rigorous testbed for advancing temporally aware multimodal systems, and will be publicly released upon acceptance.
Understanding complex surgical scenes requires recognizing multiple interdependent entities--such as instruments, actions, and targets--and maintaining their relational consistency across time. Existing surgical triplet recognition methods struggle to jointly model intra-frame label dependencies and inter-frame temporal semantics in a unified manner. To address these limitations, we propose a unified framework that integrates spatial, relational, and temporal cues for robust surgical triplet recognition. Specifically, class-specific spatial priors are first extracted through a multi-scale encoder. Then, these priors are refined by a Label Correlation Modeling module with multi-scale class activation map-guided relational extraction (MS-CAMRE), enabling the model to capture both static co-occurrence and dynamic contextual dependencies among triplet components. Furthermore, a Bidirectional Temporal-Relational Fusion Attention (BTRFA) module harmonizes temporal and relational representations to achieve coherent temporal reasoning. We also introduce a new evaluation metric, the Triplet Consistency Error Rate (TCER), which quantitatively measures the model's capability to preserve causal and semantic consistency across triplets. Extensive experiments on the CholecT45 and ProStaTD datasets show that our method achieves state-of-the-art (SOTA) performance, improving AP_ IVT by 5.1% and 7.8%, respectively. Moreover, on the TCER metric, our approach yields over 36% and 25% relative reductions on the two datasets, respectively, underscoring the effectiveness of our framework in temporal-relational co-reasoning.
Recent advances in vision-language models (VLMs) have made impressive strides in understanding spatio-temporal relationships in videos. However, when spatial information is obscured, these models struggle to capture purely temporal patterns. We introduce SpookyBench, a benchmark where information is encoded solely in temporal sequences of noise-like frames, mirroring natural phenomena from biological signaling to covert communication. Interestingly, while humans can recognize shapes, text, and patterns in these sequences with over 98% accuracy, state-of-the-art VLMs achieve 0% accuracy. This highlights a critical limitation: an over-reliance on frame-level spatial features and an inability to extract meaning from temporal cues. Overcoming this limitation will require novel architectures or training paradigms that decouple spatial dependencies from temporal processing. Our systematic analysis shows that this issue persists across model scales and architectures. We release SpookyBench to catalyze research in temporal pattern recognition and bridge the gap between human and machine video understanding. Dataset and code are available on the project website.
Annotated 3D scene data is scarce and expensive to acquire, while abundant unlabeled videos are readily available on the internet. In this paper, we demonstrate that carefully designed data engines can leverage web-curated, unlabeled videos to automatically generate training data, to facilitate end-to-end models in 3D scene understanding alongside human-annotated datasets. We identify and analyze bottlenecks in automated data generation, revealing critical factors that determine the efficiency and effectiveness of learning from unlabeled data. To validate our approach across different perception granularities, we evaluate on three tasks spanning low-level perception, i.e., 3D object detection and instance segmentation, to high-level reasoning, i.e., 3D spatial Visual Question Answering (VQA) and Vision-Lanugage Navigation (VLN). Models trained on our generated data demonstrate strong zero-shot performance and show further improvement after finetuning. This demonstrates the viability of leveraging readily available web data as a path toward more capable scene understanding systems.
Humans perceive actions through key transitions that structure actions across multiple abstraction levels, whereas machines, relying on visual features, tend to over-segment. This highlights the difficulty of enabling hierarchical reasoning in video understanding. Interestingly, we observe that lower-level visual and high-level action latent variables evolve at different rates, with low-level visual variables changing rapidly, while high-level action variables evolve more slowly, making them easier to identify. Building on this insight, we propose the Hierarchical Action Learning (HAL) model for weakly-supervised action segmentation. Our approach introduces a hierarchical causal data generation process, where high-level latent action govern the dynamics of low-level visual features. To model these varying timescales effectively, we introduce deterministic processes to align these latent variables over time. The HAL model employs a hierarchical pyramid transformer to capture both visual features and latent variables, and a sparse transition constraint is applied to enforce the slower dynamics of high-level action variables. This mechanism enhances the identification of these latent variables over time. Under mild assumptions, we prove that these latent action variables are strictly identifiable. Experimental results on several benchmarks show that the HAL model significantly outperforms existing methods for weakly-supervised action segmentation, confirming its practical effectiveness in real-world applications.
Processing long-form videos with Video Large Language Models (Video-LLMs) is computationally prohibitive. Current efficiency methods often compromise fine-grained perception through irreversible information disposal or inhibit long-range temporal modeling via rigid, predefined sparse patterns. This paper introduces AdaSpark, an adaptive sparsity framework designed to address these limitations. AdaSpark first partitions video inputs into 3D spatio-temporal cubes. It then employs two co-designed, context-aware components: (1) Adaptive Cube-Selective Attention (AdaS-Attn), which dynamically selects a subset of relevant video cubes to attend for each query token, and (2) Adaptive Token-Selective FFN (AdaS-FFN), which selectively processes only the most salient tokens within each cube. An entropy-based (Top-p) selection mechanism adaptively allocates computational resources based on input complexity. Experiments demonstrate that AdaSpark significantly reduces computational load by up to 57% FLOPs while maintaining comparable performance to dense models and preserving fine-grained, long-range dependencies, as validated on challenging hour-scale video benchmarks.
SceneScribe-1M: A Large-Scale Video Dataset with Comprehensive Geometric and Semantic Annotations
PDF ↗The convergence of 3D geometric perception and video synthesis has created an unprecedented demand for large-scale video data that is rich in both semantic and spatio-temporal information. While existing datasets have advanced either 3D understanding or video generation, a significant gap remains in providing a unified resource that supports both domains at scale. To bridge this chasm, we introduce SceneScribe-1M, a new large-scale, multi-modal video dataset. It comprises one million in-the-wild videos, each meticulously annotated with detailed textual descriptions, precise camera parameters, dense depth maps, and consistent 3D point tracks. We demonstrate the versatility and value of SceneScribe-1M by establishing benchmarks across a wide array of downstream tasks, including monocular depth estimation, scene reconstruction, and dynamic point tracking, as well as generative tasks such as text-to-video synthesis, with or without camera control. By open-sourcing SceneScribe-1M, we aim to provide a comprehensive benchmark and a catalyst for research, fostering the development of models that can both perceive the dynamic 3D world and generate controllable, realistic video content.