NeurIPS 2025 论文列表

The Arrow of Time (AoT)—time's irreversible flow shaping physical events—is fundamental to video comprehension, yet remains a significant challenge for modern large multimodal models (LMMs). Current LMMs struggle to perceive and utilize temporal directionality in video when responding to language queries, obstructing deeper temporal understanding. We tackle this deficiency by first providing a critical analysis of existing benchmarks and models. We then introduce ArrowRL, a reinforcement learning (RL)-based training strategy with an innovative reverse reward that instills AoT awareness by encouraging divergent video interpretations between forward and reversed visual frames. For rigorous evaluation, we additionally develop AoTBench, a new multi-faceted benchmark probing temporally challenging questions. Experiments show ArrowRL greatly advances temporal perception: it not only achieves substantial improvements on our challenging AoTBench but also demonstrably boosts performance on standard video question answering (VQA) benchmarks (with peak accuracy gains reaching over 20% and 10% respectively). This validates ArrowRL's effectiveness and highlights the critical need for dedicated AoT understanding in LMMs.

What does it mean for two videos to be similar? Videos may appear similar when judged by the actions they depict, yet entirely different if evaluated based on the locations where they were filmed. While humans naturally compare videos by taking different aspects into account, this ability has not been thoroughly studied and presents a challenge for models that often depend on broad global similarity scores. Large Multimodal Models (LMMs) with video understanding capabilities open new opportunities for leveraging natural language in comparative video tasks. We introduce Concept-based Video Similarity estimation (ConViS), a novel task that compares pairs of videos by computing interpretable similarity scores across a predefined set of key semantic concepts. ConViS allows for human-like reasoning about video similarity and enables new applications such as concept-conditioned video retrieval. To support this task, we also introduce ConViS-Bench, a new benchmark comprising carefully annotated video pairs spanning multiple domains. Each pair comes with concept-level similarity scores and textual descriptions of both differences and similarities. Additionally, we benchmark several state-of-the-art models on ConViS, providing insights into their alignment with human judgments. Our results reveal significant performance differences on ConViS, indicating that some concepts present greater challenges for estimating video similarity. We believe that ConViS-Bench will serve as a valuable resource for advancing research in language-driven video understanding.

Large Language Models (LLMs) are commonly pretrained on vast corpora of text without utilizing contextual metadata such as source, quality, or topic, leading to a context-free learning paradigm. While recent studies suggest that adding metadata like URL information as context (i.e., auxiliary inputs not used in the loss calculation) can improve training efficiency and downstream performance, they offer limited understanding of which types of metadata are truly effective and under what conditions. In this work, we conduct a systematic evaluation and find that not all metadata types contribute equally. Only URL context speeds up training, whereas quality scores and topic/format domain information offer no clear benefit. Furthermore, the improved downstream performances of URL conditioning emerge only when longer prompts are used at inference time. In addition, we demonstrate that context-aware pretraining enables more controllable generation than context-free pretraining, in a classifier-free guidance fashion. Although topic and format metadata do not accelerate training, they are effective for steering outputs, offering human-interpretable control over generation.

Understanding molecules is key to understanding organisms and driving advances in drug discovery, requiring interdisciplinary knowledge across chemistry and biology. Although large molecular language models have achieved notable success in task transfer, they often struggle to accurately analyze molecular features due to limited knowledge and reasoning capabilities. To address this issue, we present Mol-LLaMA, a large molecular language model that grasps the general knowledge centered on molecules and exhibits explainability and reasoning ability. To this end, we design key data types that encompass the fundamental molecular features, taking into account the essential abilities for molecular reasoning. Further, to improve molecular understanding, we propose a module that integrates complementary information from different molecular encoders, leveraging the distinct advantages of molecular representations. Our experimental results demonstrate that Mol-LLaMA is capable of comprehending the general features of molecules and providing informative responses, implying its potential as a general-purpose assistant for molecular analysis. Our project page is at https://mol-llama.github.io/.

Vision-language models (VLMs) have exhibited impressive capabilities across diverse image understanding tasks, but still struggle in settings that require reasoning over extended sequences of camera frames from a video. This limits their utility in embodied settings, which require reasoning over long frame sequences from a continuous stream of visual input at each moment of a task attempt. To address this limitation, we propose ROVER (Reasoning Over VidEo Recursively), a framework that enables the model to recursively decompose long-horizon video trajectories into segments corresponding to shorter subtasks within the trajectory. In doing so, ROVER facilitates more focused and accurate reasoning over temporally localized frame sequences without losing global context. We evaluate ROVER, implemented using an in-context learning approach, on diverse OpenX Embodiment videos and on a new dataset derived from RoboCasa that consists of 543 videos showing both expert and perturbed non-expert trajectories across 27 manipulation tasks. ROVER outperforms strong baselines across three video reasoning tasks: task progress estimation, frame-level natural language reasoning, and video question answering. We observe that, by reducing the number of frames the model reasons over at each timestep, ROVER mitigates model hallucinations, especially during unexpected or non-optimal moments of a trajectory. In addition, by enabling the implementation of a subtask-specific sliding context window, ROVER's time complexity scales linearly with video length, an asymptotic improvement over baselines.

Multi-frame video enhancement tasks aim to improve the spatial and temporal resolution and quality of video sequences by leveraging temporal information from multiple frames, which are widely used in streaming video processing, surveillance, and generation. Although numerous Transformer-based enhancement methods have achieved impressive performance, their computational and memory demands hinder deployment on edge devices. Quantization offers a practical solution by reducing the bit-width of weights and activations to improve efficiency. However, directly applying existing quantization methods to video enhancement tasks often leads to significant performance degradation and loss of fine details. This stems from two limitations: (a) inability to allocate varying representational capacity across frames, which results in suboptimal dynamic range adaptation; (b) over-reliance on full-precision teachers, which limits the learning of low-bit student models. To tackle these challenges, we propose a novel quantization method for video enhancement: Progressive Multi-Frame Quantization for Video Enhancement (PMQ-VE). This framework features a coarse-to-fine two-stage process: Backtracking-based Multi-Frame Quantization (BMFQ) and Progressive Multi-Teacher Distillation (PMTD). BMFQ utilizes a percentile-based initialization and iterative search with pruning and backtracking for robust clipping bounds. PMTD employs a progressive distillation strategy with both full-precision and multiple high-bit (INT) teachers to enhance low-bit models' capacity and quality. Extensive experiments demonstrate that our method outperforms existing approaches, achieving state-of-the-art performance across multiple tasks and benchmarks. The code will be made publicly available.

Despite the impressive progress of recent pretraining methods on multimodal tasks, existing methods are inherently biased towards either spatial modeling (e.g., CLIP) or temporal modeling (e.g., V-JEPA), limiting their joint capture of spatial details and temporal dynamics. To this end, we propose UniViT, a cluster-driven unified self-supervised learning framework that effectively captures the structured semantics of both image spatial content and video temporal dynamics through event-level and object-level clustering and discrimination. Specifically, we leverage offline clustering to generate semantic clusters across both modalities. For videos, multi-granularity event-level clustering progressively expands from single-event to structured multi-event segments, capturing coarse-to-fine temporal semantics; for images, object-level clustering captures fine-grained spatial semantics. However, while global clustering provides semantically consistent clusters, it lacks modeling of structured semantic relations (e.g., temporal event structures). To address this, we introduce a contrastive objective that leverages these semantic clusters as pseudo-label supervision to explicitly enforce structural constraints, including temporal event relations and spatial object co-occurrences, capturing structured semantics beyond categories. Meanwhile, UniViT jointly embeds structured object-level and event-level semantics into a unified representation space. Furthermore, UniViT introduces two key components: (i) Unified Rotary Position Embedding integrates relative positional embedding with frequency-aware dimension allocation to support position-invariant semantic learning and enhance the stability of structured semantics in the discrimination stage; and (ii) Variable Spatiotemporal Streams adapt to inputs of varying frame lengths, addressing the rigidity of conventional fixed-input approaches. Extensive experiments across varying model scales demonstrate that UniViT achieves state-of-the-art performance on linear probing, attentive probing, question answering, and spatial understanding tasks.

Multimodal Large Language Models (MLLMs) increasingly excel at perception,understanding, and reasoning. However, current benchmarks inadequately evaluate their ability to perform these tasks continuously in dynamic, real-world environments. To bridge this gap, we introduce RT V-Bench, a fine-grained benchmark for MLLM real-time video analysis. RTV-Bench includes three key principles: (1) Multi-Timestamp Question Answering (MTQA), where answers evolve with scene changes; (2) Hierarchical Question Structure, combining basic and advanced queries; and (3) Multi-dimensional Evaluation, assessing the ability of continuous perception, understanding, and reasoning. RTV-Bench contains 552 diverse videos (167.2 hours) and 4,631 high-quality QA pairs. We evaluated leading MLLMs, including proprietary (GPT-4o, Gemini 2.0), open-source offline (Qwen2.5-VL, VideoLLaMA3), and open-source real-time (VITA-1.5, InternLM-XComposer2.5-OmniLive) models. Experiment results show open-source real-time models largely outperform offline ones but still trail top proprietary models. Our analysis also reveals that larger model size or higher frame sampling rates do not significantly boost RTV-Bench performance, sometimes causing slight decreases.This underscores the need for better model architectures optimized for video stream processing and long sequences to advance real-time video analysis with MLLMs.

Achieving seamless viewing of high-fidelity volumetric video, comparable to 2D video experiences, remains an open challenge. Existing volumetric video compression methods either lack the flexibility to adjust quality and bitrate within a single model for efficient streaming across diverse networks and devices, or struggle with real-time decoding and rendering on lightweight mobile platforms. To address these challenges, we introduce 4DGCPro, a novel hierarchical 4D Gaussian compression framework that facilitates real-time mobile decoding and high-quality rendering via progressive volumetric video streaming in a single bitstream. Specifically, we propose a perceptually-weighted and compression-friendly hierarchical 4D Gaussian representation with motion-aware adaptive grouping to reduce temporal redundancy, preserve coherence, and enable scalable multi-level detail streaming. Furthermore, we present an end-to-end entropy-optimized training scheme, which incorporates layer-wise rate-distortion (RD) supervision and attribute-specific entropy modeling for efficient bitstream generation. Extensive experiments show that 4DGCPro enables flexible quality and variable bitrate within a single model, achieving real-time decoding and rendering on mobile devices while outperforming existing methods in RD performance across multiple datasets.

This paper presents JavisGPT, the first unified multimodal large language model (MLLM) for Joint Audio-Video (JAV) comprehension and generation. JavisGPT adopts a concise encoder–LLM–decoder architecture, featuring a SyncFusion module for spatio-temporal audio- video fusion and synchrony-aware learnable queries to bridge a pretrained JAV-DiT generator. This design enables temporally coherent video-audio understanding and generation from multimodal instructions. We design an effective three-stage training pipeline consisting of multimodal pretraining, audio-video fine-tuning, and large-scale instruction-tuning, to progressively build multimodal comprehension and generation from existing vision-language models. To support this, we further construct JavisInst-Omni, a high-quality instruction dataset with over 200K GPT-4o-curated audio-video-text dialogues that span diverse and multi-level comprehension and generation scenarios. Extensive experiments on JAV comprehension and generation benchmarks show that JavisGPT outperforms existing MLLMs, particularly in complex and temporally synchronized settings.

We present **AgMMU**, a challenging real‑world benchmark for evaluating and advancing vision-language models (VLMs) in the knowledge‑intensive domain of agriculture. Unlike prior datasets that rely on crowdsourced prompts, AgMMU is distilled from 116,231 authentic dialogues between everyday growers and USDA-authorized Cooperative Extension experts. Through a three‑stage pipeline: automated knowledge extraction, QA generation, and human verification, we construct (i) AgMMU, an evaluation set of 746 multiple‑choice questions (MCQs) and 746 open‑ended questions (OEQs), and (ii) AgBase, a development corpus of 57,079 multimodal facts covering five high-stakes agricultural topics: insect identification, species identification, disease categorization, symptom description, and management instruction. AgMMU has three key advantages: - **Authentic \& Expert‑Verified**: All facts, images, and answers originate from real farmer and gardener inquiries answered by credentialed specialists, ensuring high‑fidelity agricultural knowledge. - **Complete Development Suite**: AgMMU uniquely couples a dual‑format evaluation benchmark (MCQ and OEQ) with AgBase, a large‑scale training set, enabling both rigorous assessment and targeted improvement of VLMs. - **Knowledge‑intensive Challenge**: Our tasks demand the synergy of nuanced visual perception and domain expertise, exposing fundamental limitations of current general‑purpose models and charting a path toward robust, application‑ready agricultural AI. Benchmarking 12 leading VLMs reveals pronounced gaps in fine‑grained perception and factual grounding. Open‑sourced models trail after proprietary ones by a wide margin. Simple fine‑tuning on AgBase boosts open-sourced model performance on challenging OEQs for up to 11.6\% on average, narrowing this gap and also motivating future research to propose better strategies in knowledge extraction and distillation from AgBase. We hope AgMMU stimulates research on domain‑specific knowledge integration and trustworthy decision support in agriculture AI development.

Operating rooms (ORs) demand precise coordination among surgeons, nurses, and equipment in a fast-paced, occlusion-heavy environment, necessitating advanced perception models to enhance safety and efficiency. Existing datasets either provide partial egocentric views or sparse exocentric multi-view context, but do not explore the comprehensive combination of both. We introduce EgoExOR, the first OR dataset and accompanying benchmark to fuse first-person and third-person perspectives. Spanning 94 minutes (84,553 frames at 15 FPS) of two emulated spine procedures, Ultrasound-Guided Needle Insertion and Minimally Invasive Spine Surgery, EgoExOR integrates egocentric data (RGB, gaze, hand tracking, audio) from wearable glasses, exocentric RGB and depth from RGB-D cameras, and ultrasound imagery. Its detailed scene graph annotations, covering 36 entities and 22 relations (568,235 triplets), enable robust modeling of clinical interactions, supporting tasks like action recognition and human-centric perception. We evaluate the surgical scene graph generation performance of two adapted state-of-the-art models and offer a new baseline that explicitly leverages EgoExOR’s multimodal and multi-perspective signals. This new dataset and benchmark set a new foundation for OR perception, offering a rich, multimodal resource for next-generation clinical perception. Our code and data are available at https://github.com/ardamamur/EgoExOR.

We present LongVPO, a novel two‑stage Direct Preference Optimization framework that enables short‑context vision‑language models to robustly understand ultra‑long videos without any long‑video annotations. In Stage 1, we synthesize preference triples by anchoring questions to individual short clips, interleaving them with distractors, and applying visual‑similarity and question‑specificity filtering to mitigate positional bias and ensure unambiguous supervision. We also approximate the reference model’s scoring over long contexts by evaluating only the anchor clip, reducing computational overhead. In Stage 2, we employ a recursive captioning pipeline on long videos to generate scene-level metadata, and then use a large language model to craft multi-segment reasoning queries and dispreferred responses, aligning the model's preferences through multi-segment reasoning tasks. With only 16K synthetic examples and no costly human labels, \model{} outperforms the state‑of‑the‑art open‑source models on multiple long‑video benchmarks, while maintaining strong short‑video performance (e.g., on MVBench), offering a scalable paradigm for efficient long‑form video understanding.

Cinematography is a cornerstone of film production and appreciation, shaping mood, emotion, and narrative through visual elements such as camera movement, shot composition, and lighting. Despite recent progress in multimodal large language models (MLLMs) and video generation models, the capacity of current models to grasp and reproduce cinematographic techniques remains largely uncharted, hindered by the scarcity of expert-annotated data. To bridge this gap, we present CineTechBench, a pioneering benchmark founded on precise, manual annotation by seasoned cinematography experts across key cinematography dimensions. Our benchmark covers seven essential aspects—shot scale, shot angle, composition, camera movement, lighting, color, and focal length—and includes over 600 annotated movie images and 120 movie clips with clear cinematographic techniques. For the understanding task, we design question–answer pairs and annotated descriptions to assess MLLMs’ ability to interpret and explain cinematographic techniques. For the generation task, we assess advanced video generation models on their capacity to reconstruct cinema-quality camera movements given conditions such as textual prompts or keyframes. We conduct a large-scale evaluation on 15+ MLLMs and 5+ video generation models. Our results offer insights into the limitations of current models and future directions for cinematography understanding and generation in automatical film production and appreciation. The code and benchmark can be accessed at \url{https://github.com/PRIS-CV/CineTechBench}.

Recent advancements in foundation models for 2D vision have substantially improved the analysis of dynamic scenes from monocular videos. However, despite their strong generalization capabilities, these models often lack 3D consistency, a fundamental requirement for understanding scene geometry and motion, thereby causing severe spatial misalignment and temporal flickering in complex 3D environments. In this paper, we present Motion4D, a novel framework that addresses these challenges by integrating 2D priors from foundation models into a unified 4D Gaussian Splatting representation. Our method features a two-part iterative optimization framework: 1) Sequential optimization, which updates motion and semantic fields in consecutive stages to maintain local consistency, and 2) Global optimization, which jointly refines all attributes for long-term coherence. To enhance motion accuracy, we introduce a 3D confidence map that dynamically adjusts the motion priors, and an adaptive resampling process that inserts new Gaussians into under-represented regions based on per-pixel RGB and semantic errors. Furthermore, we enhance semantic coherence through an iterative refinement process that resolves semantic inconsistencies by alternately optimizing the semantic fields and updating prompts of SAM2. Extensive evaluations demonstrate that our Motion4D significantly outperforms both 2D foundation models and existing 3D-based approaches across diverse scene understanding tasks, including point-based tracking, video object segmentation, and novel view synthesis. Our code is available at https://hrzhou2.github.io/motion4d-web/.

Translating brain activity into meaningful visual content has long been recognized as a fundamental challenge in neuroscience and brain-computer interface research. Recent advances in EEG-based neural decoding have shown promise, yet two critical limitations remain in this area: poor generalization across subjects and constraints to specific visual tasks. We introduce NEED, the first unified framework achieving zero-shot cross-subject and cross-task generalization for EEG-based visual reconstruction. Our approach addresses three fundamental challenges: (1) cross-subject variability through an Individual Adaptation Module pretrained on multiple EEG datasets to normalize subject-specific patterns, (2) limited spatial resolution and complex temporal dynamics via a dual-pathway architecture capturing both low-level visual dynamics and high-level semantics, and (3) task specificity constraints through a unified inference mechanism adaptable to different visual domains. For video reconstruction, NEED achieves better performance than existing methods. Importantly, Our model maintains 93.7% of within-subject classification performance and 92.4% of visual reconstruction quality when generalizing to unseen subjects, while achieving an SSIM of 0.352 when transferring directly to static image reconstruction without fine-tuning, demonstrating how neural decoding can move beyond subject and task boundaries toward truly generalizable brain-computer interfaces.

Modern smartphones often feature asymmetric dual-lens systems, capturing wide-angle and ultra-wide views with complementary perspectives and details. Motion and shake can blur the wide lens, while the ultra-wide lens, despite lower resolution, retains sharper details. This natural complementarity offers valuable cues for video deblurring. However, existing methods focus mainly on single-camera inputs or symmetric stereo pairs, neglecting the cross-lens redundancy in mobile dual-camera systems. In this paper, we propose a practical video deblurring method, AsLeD-Net, which recurrently aligns and propagates temporal reference features from ultra-wide views fused with features extracted from wide-angle blurry frames. AsLeD-Net consists of two key modules: the adaptive local matching (ALM) module, which refines blurry features using $K$-nearest neighbor reference features, and the difference compensation (DC) module, which ensures spatial consistency and reduces misalignment. Additionally, AsLeD-Net uses the reference-guided motion compensation (RMC) module for temporal alignment, further improving frame-to-frame consistency in the deblurring process. We validate the effectiveness of AsLeD-Net through extensive experiments, benchmarking it against potential solutions for asymmetric lens deblurring.

Recent advances in scene understanding have leveraged multimodal large language models (MLLMs) for 3D reasoning by capitalizing on their strong 2D pretraining. However, the lack of explicit 3D data during MLLM pretraining limits 3D representation capability. In this paper, we investigate the 3D-awareness of MLLMs by evaluating multi-view correspondence and reveal a strong positive correlation between the quality of 3D-aware representation and downstream task performance. Motivated by this, we propose 3DRS, a framework that enhances MLLM 3D representation learning by introducing supervision from pretrained 3D foundation models. Our approach aligns MLLM visual features with rich 3D knowledge distilled from 3D models, effectively improving scene understanding. Extensive experiments across multiple benchmarks and MLLMs—including visual grounding, captioning, and question answering—demonstrate consistent performance gains. Code will be released to facilitate future research.

Vision-language models (VLMs) have made great strides in addressing temporal understanding tasks, which involve characterizing visual changes across a sequence of images. However, recent works have suggested that when making predictions, VLMs may rely on static feature biases, such as background or object features, rather than dynamic visual changes. Static feature biases are a type of shortcut and can contribute to systematic prediction errors on downstream tasks; as a result, identifying and characterizing error-inducing static feature biases is critical prior to real-world model deployment. Existing approaches for identifying such systematic failure modes in trained models (i) are typically designed for non-temporal settings and (ii) are challenging to evaluate in temporal settings due to the lack of quantitative evaluation frameworks. In this work, we address these challenges by introducing TRoVe, an automated approach for discovering error-inducing static feature biases learned by temporal VLMs. Given a trained VLM and an annotated validation dataset associated with a downstream classification task, TRoVe extracts candidate static features from the dataset and scores each feature by (i) the effect of the feature on classification errors as well as (ii) the extent to which the VLM relies on the feature when making predictions. In order to quantitatively evaluate TRoVe, we introduce an evaluation framework consisting of 101 trained temporal VLMs paired with ground-truth annotations for learned static feature biases. We use this framework to demonstrate that TRoVe can accurately identify error-inducing static feature biases in VLMs, achieving a 28.6% improvement over the closest baseline. Finally, we apply TRoVe to 7 off-the-shelf VLMs and 2 temporal understanding tasks, surfacing previously-unknown static feature biases and demonstrating that knowledge of learned biases can aid in improving model performance at test time. Our code is available at https://github.com/Stanford-AIMI/TRoVe.