Daily Papers

Contact locations and interaction forces are critical for contact-rich robot manipulation but are often ambiguous from vision alone. Measuring them on robots requires specialized tactile hardware and platform-specific data collection, which keeps robot tactile data scarce. Human egocentric video is far easier to scale but lacks touch. Recent predictors trained on egocentric recordings paired with tactile-glove measurements can predict this missing signal from video alone. These predictors, however, remain confined to human hands, which differ from robot hands in appearance and tactile layout. We present Ego2Contact, which learns a visual-tactile prior from human interactions and transfers it to robot hands. Its predictor fuses temporal scene context with hand-local visual features to infer bimanual tactile fields from the RGB video. Anatomical region correspondences align varying robot tactile layouts with the predictor's canonical grid, turning robot measurements into fine-tuning targets and predictions into policy inputs. Ego2Contact outperforms prior tactile predictors on human data, including unseen objects, and on robot data after adaptation. The adapted predictor provides tactile inputs for both simulated reinforcement learning and real-robot policy post-training. In simulation, policies using predicted touch outperform non-tactile baselines and recover part of the benefit of ground-truth tactile sensing. These results show that human-learned tactile priors can supply touch when direct tactile sensing is limited.

Multi-agent motion prediction is challenging because it aims to foresee the future trajectories of multiple agents (e.g. pedestrians) simultaneously in a complicated scene. Existing work addressed this challenge by either learning social spatial interactions represented by the positions of a group of pedestrians, while ignoring their temporal coherence (i.e. dependencies between different long trajectories), or by understanding the complicated scene layout (e.g. scene segmentation) to ensure safe navigation. However, unlike previous work that isolated the spatial interaction, temporal coherence, and scene layout, this paper designs a new mechanism, i.e., Dynamic and Static Context-aware Motion Predictor (DSCMP), to integrates these rich information into the long-short-term-memory (LSTM). It has three appealing benefits. (1) DSCMP models the dynamic interactions between agents by learning both their spatial positions and temporal coherence, as well as understanding the contextual scene layout.(2) Different from previous LSTM models that predict motions by propagating hidden features frame by frame, limiting the capacity to learn correlations between long trajectories, we carefully design a differentiable queue mechanism in DSCMP, which is able to explicitly memorize and learn the correlations between long trajectories. (3) DSCMP captures the context of scene by inferring latent variable, which enables multimodal predictions with meaningful semantic scene layout. Extensive experiments show that DSCMP outperforms state-of-the-art methods by large margins, such as 9.05\% and 7.62\% relative improvements on the ETH-UCY and SDD datasets respectively.

Multi-modal large language models (MLLMs) are making rapid progress toward general-purpose embodied agents. However, existing MLLMs do not reliably capture fine-grained links between low-level visual features and high-level textual semantics, leading to weak grounding and inaccurate perception. To overcome this challenge, we propose ESCA, a framework that contextualizes embodied agents by grounding their perception in spatial-temporal scene graphs. At its core is SGCLIP, a novel, open-domain, promptable foundation model for generating scene graphs that is based on CLIP. SGCLIP is trained on 87K+ open-domain videos using a neurosymbolic pipeline that aligns automatically generated captions with scene graphs produced by the model itself, eliminating the need for human-labeled annotations. We demonstrate that SGCLIP excels in both prompt-based inference and task-specific fine-tuning, achieving state-of-the-art results on scene graph generation and action localization benchmarks. ESCA with SGCLIP improves perception for embodied agents based on both open-source and commercial MLLMs, achieving state of-the-art performance across two embodied environments. Notably, ESCA significantly reduces agent perception errors and enables open-source models to surpass proprietary baselines. We release the source code for SGCLIP model training at https://github.com/video-fm/LASER and for the embodied agent at https://github.com/video-fm/ESCA.

We introduce Synthetic Visual Genome 2 (SVG2), a large-scale panoptic video scene graph dataset. SVG2 contains over 636K videos with 6.6M objects, 52.0M attributes, and 6.7M relations, providing an order-of-magnitude increase in scale and diversity over prior spatio-temporal scene graph datasets. To create SVG2, we design a fully automated pipeline that combines multi-scale panoptic segmentation, online-offline trajectory tracking with automatic new-object discovery, per-trajectory semantic parsing, and GPT-5-based spatio-temporal relation inference. Building on this resource, we train TRaSER, a video scene graph generation model. TRaSER augments VLMs with a trajectory-aligned token arrangement mechanism and new modules: an object-trajectory resampler and a temporal-window resampler to convert raw videos and panoptic trajectories into compact spatio-temporal scene graphs in a single forward pass. The temporal-window resampler binds visual tokens to short trajectory segments to preserve local motion and temporal semantics, while the object-trajectory resampler aggregates entire trajectories to maintain global context for objects. On the PVSG, VIPSeg, VidOR and SVG2 test datasets, TRaSER improves relation detection by +15 to 20%, object prediction by +30 to 40% over the strongest open-source baselines and by +13% over GPT-5, and attribute prediction by +15%. When TRaSER's generated scene graphs are sent to a VLM for video question answering, it delivers a +1.5 to 4.6% absolute accuracy gain over using video only or video augmented with Qwen2.5-VL's generated scene graphs, demonstrating the utility of explicit spatio-temporal scene graphs as an intermediate representation.

Recent end-to-end autonomous driving approaches have leveraged Vision-Language Models (VLMs) to enhance planning capabilities in complex driving scenarios. However, VLMs are inherently trained as generalist models, lacking specialized understanding of driving-specific reasoning in 3D space and time. When applied to autonomous driving, these models struggle to establish structured spatial-temporal representations that capture geometric relationships, scene context, and motion patterns critical for safe trajectory planning. To address these limitations, we propose SGDrive, a novel framework that explicitly structures the VLM's representation learning around driving-specific knowledge hierarchies. Built upon a pre-trained VLM backbone, SGDrive decomposes driving understanding into a scene-agent-goal hierarchy that mirrors human driving cognition: drivers first perceive the overall environment (scene context), then attend to safety-critical agents and their behaviors, and finally formulate short-term goals before executing actions. This hierarchical decomposition provides the structured spatial-temporal representation that generalist VLMs lack, integrating multi-level information into a compact yet comprehensive format for trajectory planning. Extensive experiments on the NAVSIM benchmark demonstrate that SGDrive achieves state-of-the-art performance among camera-only methods on both PDMS and EPDMS, validating the effectiveness of hierarchical knowledge structuring for adapting generalist VLMs to autonomous driving.

Deep saliency prediction algorithms complement the object recognition features, they typically rely on additional information, such as scene context, semantic relationships, gaze direction, and object dissimilarity. However, none of these models consider the temporal nature of gaze shifts during image observation. We introduce a novel saliency prediction model that learns to output saliency maps in sequential time intervals by exploiting human temporal attention patterns. Our approach locally modulates the saliency predictions by combining the learned temporal maps. Our experiments show that our method outperforms the state-of-the-art models, including a multi-duration saliency model, on the SALICON benchmark. Our code will be publicly available on GitHub.

Video-language models (VLMs) achieve strong multimodal understanding but remain prone to hallucinations, especially when reasoning about actions and temporal order. Existing mitigation strategies, such as textual filtering or random video perturbations, often fail to address the root cause: over-reliance on language priors rather than fine-grained visual dynamics. We propose a scalable framework for counterfactual video generation that synthesizes videos differing only in actions or temporal structure while preserving scene context. Our pipeline combines multimodal LLMs for action proposal and editing guidance with diffusion-based image and video models to generate semantic hard negatives at scale. Using this framework, we build CounterVid, a synthetic dataset of ~26k preference pairs targeting action recognition and temporal reasoning. We further introduce MixDPO, a unified Direct Preference Optimization approach that jointly leverages textual and visual preferences. Fine-tuning Qwen2.5-VL with MixDPO yields consistent improvements, notably in temporal ordering, and transfers effectively to standard video hallucination benchmarks. Code and models will be made publicly available.

Video multimodal large language models (MLLMs) have made rapid progress on general and long-form video understanding, yet their ability to preserve brief answer-critical visual evidence remains underexplored. Many practical questions are determined by momentary visual events: localized actions or state transitions that may last only a few frames. Such evidence can be skipped by sparse frame sampling, suppressed by visual-token compression, or diluted by coarse temporal aggregation, causing failures that language-side reasoning cannot reliably recover. We introduce Moment-Video, a benchmark for diagnosing the temporal fidelity of video MLLMs through momentary visual event understanding. Each question is grounded in a localized, visually observable, and sampling-sensitive event, requiring models to notice, count, describe, or reason about transient evidence rather than rely on persistent objects, global scene context, or language priors. Moment-Video contains 1,000 human-verified video-QA pairs across 7 domains and 25 fine-grained subcategories, covering four task types: Temporal Occurrence, Temporal Counting, Action Description, and Temporal Reasoning. We evaluate 33 proprietary and open-source MLLMs on Moment-Video. The best-performing model, Seed-2.0-Pro, achieves only 39.6% overall accuracy, while most open-source models remain below 25%, revealing a substantial gap in momentary visual event understanding. Diagnostic analyses show that denser frame sampling improves some models but does not eliminate the bottleneck, and longer videos introduce stronger temporal-localization challenges. These findings suggest that current video MLLMs still lack temporally faithful representations for capturing, preserving, and using brief but decisive visual evidence.

Dexterous manipulation is limited by the cost of collecting large-scale robot demonstrations. Egocentric human videos offer a scalable source of diverse manipulation behaviors, but directly using them for robot learning requires bridging two gaps: the visual gap between human and robot observations, and the action gap between human motion and robot-executable action. We propose EgoEngine, a scalable framework for transforming egocentric human manipulation videos into high-fidelity robot data. Given an egocentric RGB video, EgoEngine produces: (i) a high-fidelity robot observation video replacing human with robot while preserving scene context and temporal alignment, and (ii) a task-aligned, executable robot action trajectory under feasibility constraints. Experiments in simulation and on real robots show that EgoEngine enables scalable conversion of human videos into robot data and, to our knowledge, demonstrates the first zero-shot visuomotor dexterous policy learning from egocentric human videos without real-robot demonstrations. Project website: https://egoengine.github.io.

The rapid advancement of vision-language models (VLMs) has established a new paradigm in video anomaly detection (VAD): leveraging VLMs to simultaneously detect anomalies and provide comprehendible explanations for the decisions. Existing work in this direction often assumes the complex reasoning required for VAD exceeds the capabilities of pretrained VLMs. Consequently, these approaches either incorporate specialized reasoning modules during inference or rely on instruction tuning datasets through additional training to adapt VLMs for VAD. However, such strategies often incur substantial computational costs or data annotation overhead. To address these challenges in explainable VAD, we introduce a verbalized learning framework named VERA that enables VLMs to perform VAD without model parameter modifications. Specifically, VERA automatically decomposes the complex reasoning required for VAD into reflections on simpler, more focused guiding questions capturing distinct abnormal patterns. It treats these reflective questions as learnable parameters and optimizes them through data-driven verbal interactions between learner and optimizer VLMs, using coarsely labeled training data. During inference, VERA embeds the learned questions into model prompts to guide VLMs in generating segment-level anomaly scores, which are then refined into frame-level scores via the fusion of scene and temporal contexts. Experimental results on challenging benchmarks demonstrate that the learned questions of VERA are highly adaptable, significantly improving both detection performance and explainability of VLMs for VAD.

Camera-based 3D semantic scene completion (SSC) is pivotal for predicting complicated 3D layouts with limited 2D image observations. The existing mainstream solutions generally leverage temporal information by roughly stacking history frames to supplement the current frame, such straightforward temporal modeling inevitably diminishes valid clues and increases learning difficulty. To address this problem, we present HTCL, a novel Hierarchical Temporal Context Learning paradigm for improving camera-based semantic scene completion. The primary innovation of this work involves decomposing temporal context learning into two hierarchical steps: (a) cross-frame affinity measurement and (b) affinity-based dynamic refinement. Firstly, to separate critical relevant context from redundant information, we introduce the pattern affinity with scale-aware isolation and multiple independent learners for fine-grained contextual correspondence modeling. Subsequently, to dynamically compensate for incomplete observations, we adaptively refine the feature sampling locations based on initially identified locations with high affinity and their neighboring relevant regions. Our method ranks 1^{st} on the SemanticKITTI benchmark and even surpasses LiDAR-based methods in terms of mIoU on the OpenOccupancy benchmark. Our code is available on https://github.com/Arlo0o/HTCL.

Nowadays, navigation and ride-sharing apps have collected numerous images with spatio-temporal data. A core technology for analyzing such images, associated with spatiotemporal information, is Traffic Scene Understanding (TSU), which aims to provide a comprehensive description of the traffic scene. Unlike traditional spatio-temporal data analysis tasks, the dependence on both spatio-temporal and visual-textual data introduces distinct challenges to TSU task. However, recent research often treats TSU as a common image understanding task, ignoring the spatio-temporal information and overlooking the interrelations between different aspects of the traffic scene. To address these issues, we propose a novel SpatioTemporal Enhanced Model based on CILP (ST-CLIP) for TSU. Our model uses the classic vision-language model, CLIP, as the backbone, and designs a Spatio-temporal Context Aware Multiaspect Prompt (SCAMP) learning method to incorporate spatiotemporal information into TSU. The prompt learning method consists of two components: A dynamic spatio-temporal context representation module that extracts representation vectors of spatio-temporal data for each traffic scene image, and a bi-level ST-aware multi-aspect prompt learning module that integrates the ST-context representation vectors into word embeddings of prompts for the CLIP model. The second module also extracts low-level visual features and image-wise high-level semantic features to exploit interactive relations among different aspects of traffic scenes. To the best of our knowledge, this is the first attempt to integrate spatio-temporal information into visionlanguage models to facilitate TSU task. Experiments on two realworld datasets demonstrate superior performance in the complex scene understanding scenarios with a few-shot learning strategy.

Autoregressive video diffusion models have proved effective for world modeling and interactive scene generation, with Minecraft gameplay as a representative application. To faithfully simulate play, a model must generate natural content while exploring new scenes and preserve spatial consistency when revisiting explored areas. Under limited computation budgets, it must compress and exploit historical cues within a finite context window, which exposes a trade-off: Temporal-only memory lacks long-term spatial consistency, whereas adding spatial memory strengthens consistency but may degrade new scene generation quality when the model over-relies on insufficient spatial context. We present Memory Forcing, a learning framework that pairs training protocols with a geometry-indexed spatial memory. Hybrid Training exposes distinct gameplay regimes, guiding the model to rely on temporal memory during exploration and incorporate spatial memory for revisits. Chained Forward Training extends autoregressive training with model rollouts, where chained predictions create larger pose variations and encourage reliance on spatial memory for maintaining consistency. Point-to-Frame Retrieval efficiently retrieves history by mapping currently visible points to their source frames, while Incremental 3D Reconstruction maintains and updates an explicit 3D cache. Extensive experiments demonstrate that Memory Forcing achieves superior long-term spatial consistency and generative quality across diverse environments, while maintaining computational efficiency for extended sequences.

The robustness of semantic segmentation on edge cases of traffic scene is a vital factor for the safety of intelligent transportation. However, most of the critical scenes of traffic accidents are extremely dynamic and previously unseen, which seriously harm the performance of semantic segmentation methods. In addition, the delay of the traditional camera during high-speed driving will further reduce the contextual information in the time dimension. Therefore, we propose to extract dynamic context from event-based data with a higher temporal resolution to enhance static RGB images, even for those from traffic accidents with motion blur, collisions, deformations, overturns, etc. Moreover, in order to evaluate the segmentation performance in traffic accidents, we provide a pixel-wise annotated accident dataset, namely DADA-seg, which contains a variety of critical scenarios from traffic accidents. Our experiments indicate that event-based data can provide complementary information to stabilize semantic segmentation under adverse conditions by preserving fine-grained motion of fast-moving foreground (crash objects) in accidents. Our approach achieves +8.2% performance gain on the proposed accident dataset, exceeding more than 20 state-of-the-art semantic segmentation methods. The proposal has been demonstrated to be consistently effective for models learned on multiple source databases including Cityscapes, KITTI-360, BDD, and ApolloScape.

The monocular depth estimation task has recently revealed encouraging prospects, especially for the autonomous driving task. To tackle the ill-posed problem of 3D geometric reasoning from 2D monocular images, multi-frame monocular methods are developed to leverage the perspective correlation information from sequential temporal frames. However, moving objects such as cars and trains usually violate the static scene assumption, leading to feature inconsistency deviation and misaligned cost values, which would mislead the optimization algorithm. In this work, we present CTA-Depth, a Context-aware Temporal Attention guided network for multi-frame monocular Depth estimation. Specifically, we first apply a multi-level attention enhancement module to integrate multi-level image features to obtain an initial depth and pose estimation. Then the proposed CTA-Refiner is adopted to alternatively optimize the depth and pose. During the refinement process, context-aware temporal attention (CTA) is developed to capture the global temporal-context correlations to maintain the feature consistency and estimation integrity of moving objects. In particular, we propose a long-range geometry embedding (LGE) module to produce a long-range temporal geometry prior. Our approach achieves significant improvements over state-of-the-art approaches on three benchmark datasets.

Accurately and promptly predicting accidents among surrounding traffic agents from camera footage is crucial for the safety of autonomous vehicles (AVs). This task presents substantial challenges stemming from the unpredictable nature of traffic accidents, their long-tail distribution, the intricacies of traffic scene dynamics, and the inherently constrained field of vision of onboard cameras. To address these challenges, this study introduces a novel accident anticipation framework for AVs, termed CRASH. It seamlessly integrates five components: object detector, feature extractor, object-aware module, context-aware module, and multi-layer fusion. Specifically, we develop the object-aware module to prioritize high-risk objects in complex and ambiguous environments by calculating the spatial-temporal relationships between traffic agents. In parallel, the context-aware is also devised to extend global visual information from the temporal to the frequency domain using the Fast Fourier Transform (FFT) and capture fine-grained visual features of potential objects and broader context cues within traffic scenes. To capture a wider range of visual cues, we further propose a multi-layer fusion that dynamically computes the temporal dependencies between different scenes and iteratively updates the correlations between different visual features for accurate and timely accident prediction. Evaluated on real-world datasets--Dashcam Accident Dataset (DAD), Car Crash Dataset (CCD), and AnAn Accident Detection (A3D) datasets--our model surpasses existing top baselines in critical evaluation metrics like Average Precision (AP) and mean Time-To-Accident (mTTA). Importantly, its robustness and adaptability are particularly evident in challenging driving scenarios with missing or limited training data, demonstrating significant potential for application in real-world autonomous driving systems.

In this work, we focus on Weakly Supervised Spatio-Temporal Video Grounding (WSTVG). It is a multimodal task aimed at localizing specific subjects spatio-temporally based on textual queries without bounding box supervision. Motivated by recent advancements in multi-modal foundation models for grounding tasks, we first explore the potential of state-of-the-art object detection models for WSTVG. Despite their robust zero-shot capabilities, our adaptation reveals significant limitations, including inconsistent temporal predictions, inadequate understanding of complex queries, and challenges in adapting to difficult scenarios. We propose CoSPaL (Contextual Self-Paced Learning), a novel approach which is designed to overcome these limitations. CoSPaL integrates three core components: (1) Tubelet Phrase Grounding (TPG), which introduces spatio-temporal prediction by linking textual queries to tubelets; (2) Contextual Referral Grounding (CRG), which improves comprehension of complex queries by extracting contextual information to refine object identification over time; and (3) Self-Paced Scene Understanding (SPS), a training paradigm that progressively increases task difficulty, enabling the model to adapt to complex scenarios by transitioning from coarse to fine-grained understanding.

Long-horizon autoregressive video generation is limited by a finite context window. When an object or scene falls out of context, its fine-grained visual details may be lost and difficult to recover upon reappearance. To retain access to such visual details, we introduce MosaiChunk, a spatio-temporal memory mechanism that composes a mosaic of selected historical key-value (KV) entries across space and time. Our approach is motivated by the observation that a frozen video generator can directly consume such non-contiguous historical KV and recover the corresponding visual content. We therefore keep the generator fixed and learn only a lightweight router that determines which historical sections to include in the mosaic under a fixed active-memory budget. We further introduce RememBench, a benchmark of long-horizon revisits with prompt-driven text-to-video (T2V) and camera-driven image-to-video (I2V) splits. Our experiments show that MosaiChunk consistently improves revisit consistency over both sliding-window inference and whole-chunk retrieval under matched memory budgets, across both T2V and I2V settings.

Video scene graph generation (VidSGG) aims to capture objects and their evolving relationships in videos. This requires accounting for both the persistence of object identities and abrupt changes in relationships. Existing methods aggregate information across frames to exploit temporal context, but an emphasis on consistency can obscure these changes. As a result, a model may retain its previous prediction even after the underlying relation has changed. We refer to this failure mode as relational semantic inertia. To address this issue, we propose Relation-Transition-Aware Transformer (RTAformer), a one-stage framework that separately models entity persistence and relation changes. It injects calibrated differences between adjacent-frame relation queries to guide predicate updates, while cross-frame instance alignment during training helps reduce cross-frame representational drift. Experiments on Action Genome show improvements in recall and mean recall over existing methods, with ablation studies supporting the contributions of the proposed components and comparisons using matched backbones showing larger gains on relation-transition frames than on non-transition frames.

This paper introduces StoryAnchors, a unified framework for generating high-quality, multi-scene story frames with strong temporal consistency. The framework employs a bidirectional story generator that integrates both past and future contexts to ensure temporal consistency, character continuity, and smooth scene transitions throughout the narrative. Specific conditions are introduced to distinguish story frame generation from standard video synthesis, facilitating greater scene diversity and enhancing narrative richness. To further improve generation quality, StoryAnchors integrates Multi-Event Story Frame Labeling and Progressive Story Frame Training, enabling the model to capture both overarching narrative flow and event-level dynamics. This approach supports the creation of editable and expandable story frames, allowing for manual modifications and the generation of longer, more complex sequences. Extensive experiments show that StoryAnchors outperforms existing open-source models in key areas such as consistency, narrative coherence, and scene diversity. Its performance in narrative consistency and story richness is also on par with GPT-4o. Ultimately, StoryAnchors pushes the boundaries of story-driven frame generation, offering a scalable, flexible, and highly editable foundation for future research.

Existing multi-modal large language models (MLLMs) face significant challenges in processing long video sequences due to strict input token limitations. As a result, current video understanding approaches, especially in egocentric settings characterized by complex dynamics, frequent state changes, and moving cameras, are forced to massively subsample frames. This leads to severe loss of temporal and contextual information, constraining their ability to perform fine-grained video reasoning. In this work, we introduce a framework for egocentric video question answering (VQA) that overcomes these input constraints through Egocentric Scene Graphs (EgoSGs), i.e., temporally grounded, structured representations that capture objects, attributes, spatial relations, and interactions over time. By representing videos as compact, text-based scene graphs, our method preserves the essential visual and temporal information of the original video in a symbolic form that drastically reduces input length while maintaining semantic richness. Crucially, this enables MLLMs to reason efficiently over entire video sequences within their token budget. On HD-EPIC VQA, our method achieves state-of-the-art results, outperforming strong video-based baselines on multiple models and suggesting that structured, temporally grounded representations like EgoSGs can bridge long-form egocentric video understanding and the context limitations of today's MLLMs.

Video outpainting aims to expand the visible content of a video beyond the original frame boundaries while preserving spatial fidelity and temporal coherence across frames. Existing methods primarily rely on large-scale generative models, such as diffusion models. However, generationbased approaches suffer from implicit temporal modeling and limited spatial context. These limitations lead to intraframe and inter-frame inconsistencies, which become particularly pronounced in dynamic scenes and large outpainting scenarios. To overcome these challenges, we propose Seen-to-Scene, a novel framework that unifies propagationbased and generation-based paradigms for video outpainting. Specifically, Seen-to-Scene leverages flow-based propagation with a flow completion network pre-trained for video inpainting, which is fine-tuned in an end-to-end manner to bridge the domain gap and reconstruct coherent motion fields. To further improve the efficiency and reliability of propagation, we introduce a reference-guided latent propagation that effectively propagates source content across frames. Extensive experiments demonstrate that our method achieves superior temporal coherence and visual realism with efficient inference, surpassing even prior state-of-the-art methods that require input-specific adaptation.

We present 3DScenePrompt, a framework that generates the next video chunk from arbitrary-length input while enabling precise camera control and preserving scene consistency. Unlike methods conditioned on a single image or a short clip, we employ dual spatio-temporal conditioning that reformulates context-view referencing across the input video. Our approach conditions on both temporally adjacent frames for motion continuity and spatially adjacent content for scene consistency. However, when generating beyond temporal boundaries, directly using spatially adjacent frames would incorrectly preserve dynamic elements from the past. We address this by introducing a 3D scene memory that represents exclusively the static geometry extracted from the entire input video. To construct this memory, we leverage dynamic SLAM with our newly introduced dynamic masking strategy that explicitly separates static scene geometry from moving elements. The static scene representation can then be projected to any target viewpoint, providing geometrically consistent warped views that serve as strong 3D spatial prompts while allowing dynamic regions to evolve naturally from temporal context. This enables our model to maintain long-range spatial coherence and precise camera control without sacrificing computational efficiency or motion realism. Extensive experiments demonstrate that our framework significantly outperforms existing methods in scene consistency, camera controllability, and generation quality. Project page : https://cvlab-kaist.github.io/3DScenePrompt/

Streaming recurrent models enable efficient 3D reconstruction by maintaining persistent state representations. However, they suffer from catastrophic forgetting over long sequences due to balancing historical information with new observations. Recent methods alleviate this by deriving adaptive signals from attention perspective, but they operate on single dimensions without considering temporal and spatial consistency. To this end, we propose a training-free framework termed TTSA3R that leverages both temporal state evolution and spatial observation quality for adaptive state updates in 3D reconstruction. In particular, we devise a Temporal Adaptive Update Module that regulates update magnitude by analyzing temporal state evolution patterns. Then, a Spatial Contextual Update Module is introduced to localize spatial regions that require updates through observation-state alignment and scene dynamics. These complementary signals are finally fused to determine the state updating strategies. Extensive experiments demonstrate the effectiveness of TTSA3R in diverse 3D tasks. Moreover, our method exhibits only 1.33x error increase compared to over 4x degradation in the baseline model on extended sequences of 3D reconstruction, significantly improving long-term reconstruction stability. Our codes are available at https://github.com/anonus2357/ttsa3r.

Dynamic Scene Graph Generation (DSGG) focuses on identifying visual relationships within the spatial-temporal domain of videos. Conventional approaches often employ multi-stage pipelines, which typically consist of object detection, temporal association, and multi-relation classification. However, these methods exhibit inherent limitations due to the separation of multiple stages, and independent optimization of these sub-problems may yield sub-optimal solutions. To remedy these limitations, we propose a one-stage end-to-end framework, termed OED, which streamlines the DSGG pipeline. This framework reformulates the task as a set prediction problem and leverages pair-wise features to represent each subject-object pair within the scene graph. Moreover, another challenge of DSGG is capturing temporal dependencies, we introduce a Progressively Refined Module (PRM) for aggregating temporal context without the constraints of additional trackers or handcrafted trajectories, enabling end-to-end optimization of the network. Extensive experiments conducted on the Action Genome benchmark demonstrate the effectiveness of our design. The code and models are available at https://github.com/guanw-pku/OED.

Long video understanding (LVU) remains a core challenge in multimodal learning. Although recent vision-language models (VLMs) have made notable progress, existing benchmarks mainly focus on either fine-grained perception or coarse summarization, offering limited insight into temporal understanding over long contexts. In this work, we define a scene as a coherent segment of a video in which both visual and semantic contexts remain consistent, aligning with human perception. This leads us to a key question: can current VLMs reason effectively over long, scene-level contexts? To answer this, we introduce a new benchmark, SceneBench, designed to provide scene-level challenges. Our evaluation reveals a sharp drop in accuracy when VLMs attempt to answer scene-level questions, indicating significant forgetting of long-range context. To further validate these findings, we propose Scene Retrieval-Augmented Generation (Scene-RAG), which constructs a dynamic scene memory by retrieving and integrating relevant context across scenes. This Scene-RAG improves VLM performance by +2.50%, confirming that current models still struggle with long-context retention. We hope SceneBench will encourage future research toward VLMs with more robust, human-like video comprehension.

VLA models have emerged as a powerful paradigm for transferring semantic knowledge from web-scale data to physical robotic control. However, current single-frame architectures suffer from intrinsic limitations: temporal myopia that discards historical dynamics, reasoning gaps between high-level instructions and low-level motor commands, and inference inefficiency due to autoregressive scalar decoding. In this work, we propose MIRTH, a unified framework designed to address these challenges. MIRTH augments a pretrained VLA backbone with three key innovations: (1) dual-scale temporal memory hubs that compress long-term scene evolution and short-term motion trends into compact embeddings; (2) latent reasoning tokens optimized via a mutual-information objective carving out a semantic plan space to align multimodal context with action trajectories; and (3) a parallel action decoding scheme that replaces autoregressive generation with vector-wise prediction to maximize control throughput. Extensive evaluations on the LIBERO simulation benchmark and a real-world LeRobot platform demonstrate that MIRTH achieves state-of-the-art performance and exhibiting emergent error recovery capabilities. The codes and collected datasets are released at http://github.com/kiva12138/mirth.

Reasoning is a promising route to the generalization that autonomous driving requires in the long tail, as it can infer how the elements of a scene depend on one another and traverse those dependencies to conclusions beyond what is observed. Yet it is hard to tell whether a model's conclusions follow the scene's dependencies, because no driving representation makes them explicit enough to test against. Text-based reasoning traces lack spatio-temporal grounding, spatio-temporal scene graphs lack causal links, and reasoning annotations at scale are increasingly model-generated and hard to verify. To this end, we introduce CASCADE (Causal Spatio-Temporal Analysis of Driving Environments), which encompasses two components: (1) a structured scene representation for reasoning in driving scenes and (2) a human-annotated dataset built on it. For every actor that interacts with the ego vehicle, the CASCADE representation records frame-by-frame, for as long as the actor is visible, what action is taken, where it occurs, and how it depends on the actions and states of others. The resulting structure makes reasoning predictions machine-verifiable: they can be scored against it element by element, without relying on (M)LLM judges. The CASCADE dataset provides comprehensive human annotations for 2,066 driving clips of the PhysicalAI dataset, with over 34K elements that establish the spatio-temporal and causal context of each scene, including 8.6K time-stamped ego and agent actions, 3.7K causal links and 2.9K potential influences, and 6.1K annotations for agents, objects, traffic lights, and environments. Being entirely human-annotated, CASCADE provides the reference for this comparison: benchmarking the reasoning abilities of Physical AI models, and verifying the quality of automatically generated reasoning labels. The CASCADE dataset is available at https://huggingface.co/datasets/nvidia/cascade.

Understanding the motion states of the surrounding environment is critical for safe autonomous driving. These motion states can be accurately derived from scene flow, which captures the three-dimensional motion field of points. Existing LiDAR scene flow methods extract spatial features from each point cloud and then fuse them channel-wise, resulting in the implicit extraction of spatio-temporal features. Furthermore, they utilize 2D Bird's Eye View and process only two frames, missing crucial spatial information along the Z-axis and the broader temporal context, leading to suboptimal performance. To address these limitations, we propose Flow4D, which temporally fuses multiple point clouds after the 3D intra-voxel feature encoder, enabling more explicit extraction of spatio-temporal features through a 4D voxel network. However, while using 4D convolution improves performance, it significantly increases the computational load. For further efficiency, we introduce the Spatio-Temporal Decomposition Block (STDB), which combines 3D and 1D convolutions instead of using heavy 4D convolution. In addition, Flow4D further improves performance by using five frames to take advantage of richer temporal information. As a result, the proposed method achieves a 45.9% higher performance compared to the state-of-the-art while running in real-time, and won 1st place in the 2024 Argoverse 2 Scene Flow Challenge. The code is available at https://github.com/dgist-cvlab/Flow4D.

This paper proposes dynamic human group detection in videos. For detecting complex groups, not only the local appearance features of in-group members but also the global context of the scene are important. Such local and global appearance features in each frame are extracted using a Vision-Language Model (VLM) augmented for group detection in our method. For further improvement, the group structure should be consistent over time. While previous methods are stabilized on the assumption that groups are not changed in a video, our method detects dynamically changing groups by global optimization using a graph with all frames' groupness probabilities estimated by our groupness-augmented CLIP features. Our experimental results demonstrate that our method outperforms state-of-the-art group detection methods on public datasets. Code: https://github.com/irajisamurai/VLM-GroupDetection.git

Minute-scale cinematic video generation is a central challenge for generative video models. Existing paradigms address only fragments of this challenge: single-shot extrapolation preserves an anchor but lacks cinematic structure, while multi-shot storytelling imposes structure yet remains free to invent its visual states rather than continue an observed one. We define Multi-Shot Video Extrapolation (MSVE), a task that extends an observed frame or clip into a sequence of cinematically structured shots while preserving anchor state and advancing narrative intent. This setting operates under the finite per-call generation budget of short-video models. We identify three coupled bottlenecks: (1) global planners over-specify unsupported details from full screenplays; (2) shot-level prompts dilute task-relevant state when carrying the complete story; and (3) temporal chaining turns generated frames into a lossy memory in which identity, scene, object, and action state decay. MSVE reveals that long-video failure is not merely a limitation of context length, but a failure of context allocation. We propose Recursive Context Allocation (ReCA), an inference-time framework that allocates context hierarchically across planning and generation. ReCA recursively decomposes MSVE into context-bounded subproblems, invokes frozen generators at leaf nodes, and propagates structured state updates across time. To evaluate this setting, we further propose MSVE-Bench and NB-Q, a source-grounded protocol with prompts purpose-built for 3 to 5 minute long-video generation, a regime not addressed by existing short-clip benchmarks. Compared to previous methods, ReCA improves average normalized score by 8 to 16 percent over the strongest competing controller and improves multi-shot consistency metrics by 28 to 43 percent. View the project page at https://reca.vmv.re.

Driving world models serve as a pivotal technology for autonomous driving by simulating environmental dynamics. However, existing approaches predominantly focus on future scene generation, often overlooking comprehensive 3D scene understanding. Conversely, while Large Language Models (LLMs) demonstrate impressive reasoning capabilities, they lack the capacity to predict future geometric evolution, creating a significant disparity between semantic interpretation and physical simulation. To bridge this gap, we propose HERMES++, a unified driving world model that integrates 3D scene understanding and future geometry prediction within a single framework. Our approach addresses the distinct requirements of these tasks through synergistic designs. First, a BEV representation consolidates multi-view spatial information into a structure compatible with LLMs. Second, we introduce LLM-enhanced world queries to facilitate knowledge transfer from the understanding branch. Third, a Current-to-Future Link is designed to bridge the temporal gap, conditioning geometric evolution on semantic context. Finally, to enforce structural integrity, we employ a Joint Geometric Optimization strategy that integrates explicit geometric constraints with implicit latent regularization to align internal representations with geometry-aware priors. Extensive evaluations on multiple benchmarks validate the effectiveness of our method. HERMES++ achieves strong performance, outperforming specialist approaches in both future point cloud prediction and 3D scene understanding tasks. The model and code will be publicly released at https://github.com/H-EmbodVis/HERMESV2.

We introduce the task of arbitrary spatio-temporal video completion, where a video is generated from arbitrary, user-specified patches placed at any spatial location and timestamp, akin to painting on a video canvas. This flexible formulation naturally unifies many existing controllable video generation tasks--including first-frame image-to-video, inpainting, extension, and interpolation--under a single, cohesive paradigm. Realizing this vision, however, faces a fundamental obstacle in modern latent video diffusion models: the temporal ambiguity introduced by causal VAEs, where multiple pixel frames are compressed into a single latent representation, making precise frame-level conditioning structurally difficult. We address this challenge with VideoCanvas, a novel framework that adapts the In-Context Conditioning (ICC) paradigm to this fine-grained control task with zero new parameters. We propose a hybrid conditioning strategy that decouples spatial and temporal control: spatial placement is handled via zero-padding, while temporal alignment is achieved through Temporal RoPE Interpolation, which assigns each condition a continuous fractional position within the latent sequence. This resolves the VAE's temporal ambiguity and enables pixel-frame-aware control on a frozen backbone. To evaluate this new capability, we develop VideoCanvasBench, the first benchmark for arbitrary spatio-temporal video completion, covering both intra-scene fidelity and inter-scene creativity. Experiments demonstrate that VideoCanvas significantly outperforms existing conditioning paradigms, establishing a new state of the art in flexible and unified video generation.

Video Temporal Grounding (VTG) localizes the temporal boundaries of query-relevant moments in long, untrimmed videos, making video-language-model prohibitively expensive. While recent training-free token pruning has shown success in video question answering, naively applying these objectives to VTG causes drastic degradation, as VTG crucially depends on boundary-sensitive evidence and cross-frame reasoning chains. We therefore identify two VTG-specific pruning principles: evidence retention, which keeps query-critical patches especially around event boundaries, and connectivity strength, which preserves cross-frame connectivity for long-range evidence aggregation. Building on these insights, we propose SemVID, a training-free pruning framework that constructs a compact yet coherent token subset with complementary semantic roles. SemVID first allocates per-frame budgets by balancing query relevance and inter-frame variation to avoid over-pruned segments, and then selects three types of tokens: object tokens for diverse query-critical evidence, motion tokens to capture meaningful transitions and serve as cross-frame relays, and context tokens for scene continuity. Extensive experiments show that SemVID achieves a strong accuracy-efficiency trade-off, retaining up to 95.4% mIoU with only 12.5% visual tokens and delivering up to a 5.8x prefill speedup, consistently outperforming prior methods under the same budgets. Our code is available at https://github.com/JiaqiLi404/SemVID

Multimodal large language models have achieved strong performance across diverse vision-language tasks, yet their capabilities in UAV scenarios remain insufficiently explored. Recent UAV-oriented benchmarks have begun to evaluate MLLMs in aerial scenarios, but they typically focus on scene understanding, event recognition, or navigation completion, rather than jointly assessing the dual-cognition capability required for UAV agents: reasoning about both the UAV's own state and the external environment in multiview spatio-temporal contexts. To address this gap, we present UAV-DualCog, a benchmark for aerial multiview spatio-temporal reasoning built on this dual-cognition perspective. UAV-DualCog includes both image and video tasks to jointly evaluate self-state and environment-state reasoning, while requiring spatial or temporal grounding beyond discrete answer prediction. We also develop an automated pipeline that constructs data from scene-level semantic point clouds, yielding a scalable benchmark with diverse scenes, hundreds of landmarks, and thousands of QA samples. Extensive evaluations show that current MLLMs remain far from reliable in UAV dual cognition. Self-state reasoning, viewpoint transformation, precise spatial grounding, and temporal interval localization are persistent bottlenecks, and additional validation with thinking/frontier models and a human baseline confirms that the benchmark is understandable to humans but challenging for existing models. We further construct UAV-DualCog-Train from disjoint scenes and show through a lightweight optimization probe that it provides useful structured supervision, suggesting its value not only as an evaluation benchmark but also as a data resource for advancing MLLM-based UAV agents. Project website and supplementary materials: https://uav-dualcog.lozumi.com

Existing talking avatar methods typically adopt an image-to-video pipeline conditioned on a static reference image within the same scene as the target generation. This restricted, single-view perspective lacks sufficient temporal and expression cues, limiting the ability to synthesize high-fidelity talking avatars in customized backgrounds. To this end, we introduce Talking Avatar generation from Video Reference (TAVR), a novel framework that shifts the paradigm by leveraging cross-scene video inputs. To effectively process these extended temporal contexts and bridge cross-scene domain gaps, TAVR integrates a token selection module alongside a comprehensive three-stage training scheme. Specifically, same-scene video pretraining establishes foundational appearance copying, which is subsequently expanded by cross-scene reference fine-tuning for robust cross-scene adaptation. Finally, task-specific reinforcement learning aligns the generated outputs with identity-based rewards to maximize identity similarity. To systematically evaluate cross-scene robustness, we construct a new benchmark comprising 158 carefully curated cross-scene video pairs. Extensive experiments show that TAVR benefits from flexible inference-time video referencing and consistently surpasses existing baselines both quantitatively and qualitatively. This work has been deployed to production. For more related research, please visit https://www.heygen.com/research{HeyGen Research} and https://www.heygen.com/research/avatar-v-model{HeyGen Avatar-V}.

Emerging world models autoregressively generate video frames in response to actions, such as camera movements and text prompts, among other control signals. Due to limited temporal context window sizes, these models often struggle to maintain scene consistency during revisits, leading to severe forgetting of previously generated environments. Inspired by the mechanisms of human memory, we introduce a novel framework to enhancing long-term consistency of video world models through a geometry-grounded long-term spatial memory. Our framework includes mechanisms to store and retrieve information from the long-term spatial memory and we curate custom datasets to train and evaluate world models with explicitly stored 3D memory mechanisms. Our evaluations show improved quality, consistency, and context length compared to relevant baselines, paving the way towards long-term consistent world generation.

Video world models enable interactive exploration of dynamic environments, yet struggle to respect prior observations over long horizons and across viewpoints. We present WorldCrafter, a video world model that learns a camera-queryable implicit 3D-aware memory for this purpose. The key insight is to let the requested viewpoint shape how multi-view evidence is compressed into the video generator's limited token budget. Trained jointly with the video generator, a memory encoder and pose-conditioned readout module integrate historical observations into a fixed set of target view-specific tokens before denoising, without explicit depth-based correspondences. By combining this memory with recent temporal context and few-step distillation, WorldCrafter enables streaming scene exploration from a single input image or text prompt. Experiments across static and dynamic scenes show substantial gains in long-horizon consistency and camera-control accuracy while preserving visual quality during minute-scale exploration.

Dynamic scene reconstruction from monocular video remains a fundamental challenge in computer vision. Existing feed-forward methods predict 3D Gaussians pixel-wise for each frame, suffering from duplicated Gaussians and view-dependent biases that hinder effective learning of scene motion. We present C4G, a feed-forward 4D reconstruction framework built upon a compact set of timestamp-conditioned learnable Gaussian query tokens. Each token aggregates corresponding features across the full temporal context and decodes a 3D Gaussian whose position is modulated by the target timestamp, enabling globally coherent motion modeling without per-scene optimization. To capture fine-grained details, we further introduce a video diffusion model-based rendering enhancement module. Since our framework effectively aggregates features into Gaussians, we extend this capability to feature lifting, producing a 4D feature field that supports point tracking and dynamic scene understanding. C4G achieves strong novel-view synthesis performance using significantly fewer Gaussians and without requiring camera poses, while exhibiting stronger motion modeling and robustness to large temporal gaps.

While head-mounted devices are becoming more compact, they provide egocentric views with significant self-occlusions of the device user. Hence, existing methods often fail to accurately estimate complex 3D poses from egocentric views. In this work, we propose a new transformer-based framework to improve egocentric stereo 3D human pose estimation, which leverages the scene information and temporal context of egocentric stereo videos. Specifically, we utilize 1) depth features from our 3D scene reconstruction module with uniformly sampled windows of egocentric stereo frames, and 2) human joint queries enhanced by temporal features of the video inputs. Our method is able to accurately estimate human poses even in challenging scenarios, such as crouching and sitting. Furthermore, we introduce two new benchmark datasets, i.e., UnrealEgo2 and UnrealEgo-RW (RealWorld). The proposed datasets offer a much larger number of egocentric stereo views with a wider variety of human motions than the existing datasets, allowing comprehensive evaluation of existing and upcoming methods. Our extensive experiments show that the proposed approach significantly outperforms previous methods. We will release UnrealEgo2, UnrealEgo-RW, and trained models on our project page.

Real-world spatial intelligence requires reasoning over a continuous and evolving 3D world, yet existing VLMs and tool-augmented agents largely remain tied to static, stateless inference from isolated visual observations. We introduce \textsc{S-Agent}, a spatial tool-use agentic paradigm for understanding and reasoning over continuous multi-view images and videos. By formulating spatial reasoning as spatio-temporal evidence accumulation rather than isolated frame-level prediction, S-Agent reshapes spatial perception into scene-centric understanding beyond frame-centric recognition. Specifically, S-Agent casts the VLM as a semantic planner that decides what evidence is needed, while a hierarchy of spatial tools and experts grounds objects in 2D, lifts them into 3D geometric evidence, and aggregates this evidence into high-level spatial knowledge (e.g., counting, measurement, orientation, and relative position). Additionally, a temporal memory mechanism, including Scene Memory for maintaining the evolving scene state and Agent Memory for accumulating reasoning context, enables evidence integration across frames and reasoning steps. Comprehensive experiments on multi-view and video spatial reasoning benchmarks show that S-Agent consistently improves both open-source and closed-source VLMs in a training-free manner. Beyond inference-time augmentation, supervised fine-tuning (SFT) on S-Agent-generated spatial trajectories S-300K yields S-Agent-8B, a compact spatial agent that significantly surpasses similar-scale baselines (e.g., Qwen3-VL-8B) and performs comparably to advanced closed-source models (e.g., GPT-5.4 and Gemini 3).

Semantic segmentation of LiDAR point clouds has been widely studied in recent years, with most existing methods focusing on tackling this task using a single scan of the environment. However, leveraging the temporal stream of observations can provide very rich contextual information on regions of the scene with poor visibility (e.g., occlusions) or sparse observations (e.g., at long range), and can help reduce redundant computation frame after frame. In this paper, we tackle the challenge of exploiting the information from the past frames to improve the predictions of the current frame in an online fashion. To address this challenge, we propose a novel framework for semantic segmentation of a temporal sequence of LiDAR point clouds that utilizes a memory network to store, update and retrieve past information. Our framework also includes a regularizer that penalizes prediction variations in the neighborhood of the point cloud. Prior works have attempted to incorporate memory in range view representations for semantic segmentation, but these methods fail to handle occlusions and the range view representation of the scene changes drastically as agents nearby move. Our proposed framework overcomes these limitations by building a sparse 3D latent representation of the surroundings. We evaluate our method on SemanticKITTI, nuScenes, and PandaSet. Our experiments demonstrate the effectiveness of the proposed framework compared to the state-of-the-art.

Leveraging Vision Foundation Models (VFMs) for camera-to-LiDAR knowledge distillation offers a promising solution to the scarcity of annotated data needed to represent the immense geometric and kinematic diversity of real-world autonomous driving (AD). However, current approaches typically treat VFMs as black-box teachers, relying exclusively on frame-wise feature similarity. Consequently, they do not fully exploit the teacher's layer-wise semantic structure and global context, as well as the rich spatiotemporal information inherent in LiDAR sequences. We propose HilDA, a self-supervised pretraining framework for LiDAR backbones that better captures the semantic what and geometric where needed for driving tasks. HilDA combines hierarchical distillation comprising multi-layer distillation for progressive semantic alignment and global context distillation for scene-level semantics, with a temporal occupancy diffusion objective promoting spatiotemporal consistency. Models pre-trained with HilDA achieve state-of-the-art results on cross-modal distillation benchmarks and outperform models trained via prior distillation approaches on 3D object detection, scene flow, and semantic occupancy prediction. Code available at: https://maxiuw.github.io/hilda.

Translating novels into films poses a grand challenge for generative artificial intelligence, requiring conversion of abstract literary prose into long-form, multi-scene visual narratives. While current video generation models excel at short, single-scene clips within narrow temporal and spatial contexts, novel-to-film generation operates in a more complex regime, demanding long-duration content across diverse scenes with dynamically evolving entity states. To address this, we formalize novel-to-film generation as dynamic cinematic world modeling, decomposed into two phases: construction, which grounds abstract, underspecified literary narratives into concrete, stateful, and persistent world entities; and evolution, which governs how these entities dynamically update under plot progression to maintain causal consistency across scenes. We propose FilmWorld, an end-to-end agentic system where two groups of specialized agents collaborate to instantiate these phases. Construction-side agents perform narrative structured translation, world entity state modeling with visual anchoring, and state-driven shot planning, progressively projecting literary language into a cinematic blueprint. Evolution-side agents perform state-anchored visual generation, cross-shot dynamic state propagation, and closed-loop state verification to maintain causal consistency and visual coherence. To address the evaluation gap in long-form generation, we introduce FilmEval, a systematic evaluation framework that couples a difficulty-graded benchmark of 15 representative novels with an automated protocol of nine objective metrics spanning three dimensions: cinematic presentation, film consistency, and novel fidelity. Experiments demonstrate that FilmWorld consistently outperforms state-of-the-art video generation agent systems, with particularly pronounced improvements in narrative fidelity and cross-scene consistency.

We propose an approach to self-supervised representation learning based on maximizing mutual information between features extracted from multiple views of a shared context. For example, one could produce multiple views of a local spatio-temporal context by observing it from different locations (e.g., camera positions within a scene), and via different modalities (e.g., tactile, auditory, or visual). Or, an ImageNet image could provide a context from which one produces multiple views by repeatedly applying data augmentation. Maximizing mutual information between features extracted from these views requires capturing information about high-level factors whose influence spans multiple views -- e.g., presence of certain objects or occurrence of certain events. Following our proposed approach, we develop a model which learns image representations that significantly outperform prior methods on the tasks we consider. Most notably, using self-supervised learning, our model learns representations which achieve 68.1% accuracy on ImageNet using standard linear evaluation. This beats prior results by over 12% and concurrent results by 7%. When we extend our model to use mixture-based representations, segmentation behaviour emerges as a natural side-effect. Our code is available online: https://github.com/Philip-Bachman/amdim-public.

The Automated Audio Captioning (AAC) task asks models to generate natural language descriptions of an audio input. Evaluating these machine-generated audio captions is a complex task that requires considering diverse factors, among them, auditory scene understanding, sound-object inference, temporal coherence, and the environmental context of the scene. While current methods focus on specific aspects, they often fail to provide an overall score that aligns well with human judgment. In this work, we propose CLAIR-A, a simple and flexible method that leverages the zero-shot capabilities of large language models (LLMs) to evaluate candidate audio captions by directly asking LLMs for a semantic distance score. In our evaluations, CLAIR-A better predicts human judgements of quality compared to traditional metrics, with a 5.8% relative accuracy improvement compared to the domain-specific FENSE metric and up to 11% over the best general-purpose measure on the Clotho-Eval dataset. Moreover, CLAIR-A offers more transparency by allowing the language model to explain the reasoning behind its scores, with these explanations rated up to 30% better by human evaluators than those provided by baseline methods. CLAIR-A is made publicly available at https://github.com/DavidMChan/clair-a.

Remote sensing change detection aims to compare two or more images recorded for the same area but taken at different time stamps to quantitatively and qualitatively assess changes in geographical entities and environmental factors. Mainstream models usually built on pixel-by-pixel change detection paradigms, which cannot tolerate the diversity of changes due to complex scenes and variation in imaging conditions. To address this shortcoming, this paper rethinks the change detection with the mask view, and further proposes the corresponding: 1) meta-architecture CDMask and 2) instance network CDMaskFormer. Components of CDMask include Siamese backbone, change extractor, pixel decoder, transformer decoder and normalized detector, which ensures the proper functioning of the mask detection paradigm. Since the change query can be adaptively updated based on the bi-temporal feature content, the proposed CDMask can adapt to different latent data distributions, thus accurately identifying regions of interest changes in complex scenarios. Consequently, we further propose the instance network CDMaskFormer customized for the change detection task, which includes: (i) a Spatial-temporal convolutional attention-based instantiated change extractor to capture spatio-temporal context simultaneously with lightweight operations; and (ii) a scene-guided axial attention-instantiated transformer decoder to extract more spatial details. State-of-the-art performance of CDMaskFormer is achieved on five benchmark datasets with a satisfactory efficiency-accuracy trade-off. Code is available at https://github.com/xwmaxwma/rschange.

Multimodal Large Language Models (MLLMs) have demonstrated significant advances in visual understanding tasks involving both images and videos. However, their capacity to comprehend human-centric video data remains underexplored, primarily due to the absence of comprehensive and high-quality evaluation benchmarks. Existing human-centric benchmarks predominantly emphasize video generation quality and action recognition, while overlooking essential perceptual and cognitive abilities required in human-centered scenarios. Furthermore, they are often limited by single-question paradigms and overly simplistic evaluation metrics. To address above limitations, we propose a modern HV-MMBench, a rigorously curated benchmark designed to provide a more holistic evaluation of MLLMs in human-centric video understanding. Compared to existing human-centric video benchmarks, our work offers the following key features: (1) Diverse evaluation dimensions: HV-MMBench encompasses 13 tasks, ranging from basic attribute perception (e.g., age estimation, emotion recognition) to advanced cognitive reasoning (e.g., social relationship prediction, intention prediction), enabling comprehensive assessment of model capabilities; (2) Varied data types: The benchmark includes multiple-choice, fill-in-blank, true/false, and open-ended question formats, combined with diverse evaluation metrics, to more accurately and robustly reflect model performance; (3) Multi-domain video coverage: The benchmark spans 50 distinct visual scenarios, enabling comprehensive evaluation across fine-grained scene variations; (4) Temporal coverage: The benchmark covers videos from short-term (10 seconds) to long-term (up to 30min) durations, supporting systematic analysis of models temporal reasoning abilities across diverse contextual lengths.

We address the problem of retrieving a specific moment from an untrimmed video by natural language. It is a challenging problem because a target moment may take place in the context of other temporal moments in the untrimmed video. Existing methods cannot tackle this challenge well since they do not fully consider the temporal contexts between temporal moments. In this paper, we model the temporal context between video moments by a set of predefined two-dimensional maps under different temporal scales. For each map, one dimension indicates the starting time of a moment and the other indicates the duration. These 2D temporal maps can cover diverse video moments with different lengths, while representing their adjacent contexts at different temporal scales. Based on the 2D temporal maps, we propose a Multi-Scale Temporal Adjacent Network (MS-2D-TAN), a single-shot framework for moment localization. It is capable of encoding the adjacent temporal contexts at each scale, while learning discriminative features for matching video moments with referring expressions. We evaluate the proposed MS-2D-TAN on three challenging benchmarks, i.e., Charades-STA, ActivityNet Captions, and TACoS, where our MS-2D-TAN outperforms the state of the art.

In recent years, audio generation has made significant progress in tasks such as text-to-speech (TTS), text-to-audio (TTA) and text-to-music (TTM). However, generating long-form and controllable audio from complex audio scene descriptions remains a significant challenge, as such scenes often require coordinated speech, sound effects, music, songs, temporal structure, and post-production. In this work, we introduce Audio-Oscar, a multi-agent framework for generating audio from complex descriptions. Audio-Oscar coordinates a set of specialist agents, each responsible for a different aspect of the audio scene, including character modeling and voice design, speech generation, fine-grained timeline planning, model selection, non-speech generation, and audio post-production. Audio-Oscar further incorporates feedback-driven refinement. In addition, to address the lack of suitable benchmarks for evaluating audio generation from complex audio scene descriptions, we construct ASG-Bench, an Audio Scene Generation Benchmark containing both scene descriptions paired with reference audio and text-only scene descriptions. Each scene is annotated with target audio events and temporal statements to evaluate whether the generated audio faithfully realizes the required scene content and temporal structure. Experimental results show that Audio-Oscar can effectively generate audio that matches complex scene descriptions. Project samples are available at https://audiooscar.github.io/. Our code is available at https://github.com/ziye26/Audio-Oscar.

This paper aims to address the challenge of reconstructing long volumetric videos from multi-view RGB videos. Recent dynamic view synthesis methods leverage powerful 4D representations, like feature grids or point cloud sequences, to achieve high-quality rendering results. However, they are typically limited to short (1~2s) video clips and often suffer from large memory footprints when dealing with longer videos. To solve this issue, we propose a novel 4D representation, named Temporal Gaussian Hierarchy, to compactly model long volumetric videos. Our key observation is that there are generally various degrees of temporal redundancy in dynamic scenes, which consist of areas changing at different speeds. Motivated by this, our approach builds a multi-level hierarchy of 4D Gaussian primitives, where each level separately describes scene regions with different degrees of content change, and adaptively shares Gaussian primitives to represent unchanged scene content over different temporal segments, thus effectively reducing the number of Gaussian primitives. In addition, the tree-like structure of the Gaussian hierarchy allows us to efficiently represent the scene at a particular moment with a subset of Gaussian primitives, leading to nearly constant GPU memory usage during the training or rendering regardless of the video length. Extensive experimental results demonstrate the superiority of our method over alternative methods in terms of training cost, rendering speed, and storage usage. To our knowledge, this work is the first approach capable of efficiently handling minutes of volumetric video data while maintaining state-of-the-art rendering quality. Our project page is available at: https://zju3dv.github.io/longvolcap.

In egocentric videos, actions occur in quick succession. We capitalise on the action's temporal context and propose a method that learns to attend to surrounding actions in order to improve recognition performance. To incorporate the temporal context, we propose a transformer-based multimodal model that ingests video and audio as input modalities, with an explicit language model providing action sequence context to enhance the predictions. We test our approach on EPIC-KITCHENS and EGTEA datasets reporting state-of-the-art performance. Our ablations showcase the advantage of utilising temporal context as well as incorporating audio input modality and language model to rescore predictions. Code and models at: https://github.com/ekazakos/MTCN.

Video games provide a scalable source of training data for video world models, offering diverse environments, complex interactions, and abundant in-the-wild gameplay videos. However, raw gameplay footage entangles the game world with screen-space interfaces, introducing game-specific biases and irrelevant dynamics that hinder world-model training. To address this problem, we introduce GameUI-Taxonomy and G2WEngine, a full-stack framework that formalizes gameplay UI grounding and removal. G2WEngine automatically extracts reusable UI assets from real gameplay videos and synthesizes temporally coherent UI overlays on clean footage. Using this engine, we construct Game2World, comprising 96K synthetic paired videos with precise reconstruction targets and 1,079 in-the-wild clips from 303 games for realistic evaluation. Its asset library contains 5,132 verified UI elements across 21 taxonomy categories, collected from 1,010 representative gameplay frames. Based on Game2World, we propose GameCleaner, a mask-free gameplay UI removal model that combines multimodal semantic understanding with video editing capabilities. Unlike mask-based methods, GameCleaner directly identifies and removes diverse HUD elements while preserving the underlying scene content and temporal dynamics. In a controlled pilot, world models trained on UI-free gameplay improve overall VideoReward by 6.83% over those trained on UI-overlaid data. On UI-removal evaluation, GameCleaner achieves an average AAR of 95.36 on synthetic videos, outperforming the strongest temporal mask baseline by 57.3%, and obtains the best in-the-wild AAR of 80.05 with 99.8 background preservation. These results demonstrate the scalable potential of transforming Internet gameplay videos into high-quality world-model training data. Code, dataset, and model will be available at https://github.com/Dongping-Chen/Game2World.

Video object detection on edge devices runs computationally expensive detectors over long frame streams, causing high energy consumption and sustained GPU utilization. Although consecutive frames are highly redundant, naive frame skipping is content-blind: it skips during critical moments such as object entry, occlusion recovery, and abrupt motion, degrading detection quality. We present Albireo, a detector-agnostic, codec-free adaptive inference framework that wraps off-the-shelf detectors and decides when detector invocation can be safely skipped based on scene content and per-object temporal state, requiring no detector modification or retraining. Albireo maintains a 10-dimensional Kalman filter (KF) per active object and invokes the detector only when prediction uncertainty exceeds a threshold; on skipped frames, boxes are predicted from the KF state at near-zero GPU cost. A KF-based rescue mechanism preserves confirmed objects through brief detector misses to prevent output fragmentation, while a lightweight empty-scene screen avoids detector calls on objectless frames. We evaluate Albireo on the BDD100K MOT validation split with three architecturally distinct detectors (YOLO11x, YOLO26x, RF-DETR-Large) on two NVIDIA Jetson platforms (AGX Thor, AGX Orin). Across all configurations, Albireo keeps AP@50 within +/-1.2 pp of per-frame inference while reducing total energy by 12.1-17.6%. On YOLO26x, it improves AP@50 by +0.8 pp while reducing energy by 17.6% (Thor) and 14.4% (Orin) and per-frame energy-delay product by 24.9% and 26.1%, respectively. Thus, the default operating point improves accuracy, energy, and latency together. In contrast, FixedSkip-2, a fixed-interval baseline with a 50% skip rate, loses 8.6 pp AP@50. Source code, evaluation pipeline, and per-clip results are available at https://github.com/amirtaherin/albireo

In this work, we aim to reconstruct a time-varying 3D model, capable of rendering photo-realistic renderings with independent control of viewpoint, illumination, and time, from Internet photos of large-scale landmarks. The core challenges are twofold. First, different types of temporal changes, such as illumination and changes to the underlying scene itself (such as replacing one graffiti artwork with another) are entangled together in the imagery. Second, scene-level temporal changes are often discrete and sporadic over time, rather than continuous. To tackle these problems, we propose a new scene representation equipped with a novel temporal step function encoding method that can model discrete scene-level content changes as piece-wise constant functions over time. Specifically, we represent the scene as a space-time radiance field with a per-image illumination embedding, where temporally-varying scene changes are encoded using a set of learned step functions. To facilitate our task of chronology reconstruction from Internet imagery, we also collect a new dataset of four scenes that exhibit various changes over time. We demonstrate that our method exhibits state-of-the-art view synthesis results on this dataset, while achieving independent control of viewpoint, time, and illumination.

We propose a novel framework for video understanding, called Temporally Contextualized CLIP (TC-CLIP), which leverages essential temporal information through global interactions in a spatio-temporal domain within a video. To be specific, we introduce Temporal Contextualization (TC), a layer-wise temporal information infusion mechanism for videos, which 1) extracts core information from each frame, 2) connects relevant information across frames for the summarization into context tokens, and 3) leverages the context tokens for feature encoding. Furthermore, the Video-conditional Prompting (VP) module processes context tokens to generate informative prompts in the text modality. Extensive experiments in zero-shot, few-shot, base-to-novel, and fully-supervised action recognition validate the effectiveness of our model. Ablation studies for TC and VP support our design choices. Our project page with the source code is available at https://github.com/naver-ai/tc-clip

Modeling long-term context in videos is crucial for many fine-grained tasks including temporal action segmentation. An interesting question that is still open is how much long-term temporal context is needed for optimal performance. While transformers can model the long-term context of a video, this becomes computationally prohibitive for long videos. Recent works on temporal action segmentation thus combine temporal convolutional networks with self-attentions that are computed only for a local temporal window. While these approaches show good results, their performance is limited by their inability to capture the full context of a video. In this work, we try to answer how much long-term temporal context is required for temporal action segmentation by introducing a transformer-based model that leverages sparse attention to capture the full context of a video. We compare our model with the current state of the art on three datasets for temporal action segmentation, namely 50Salads, Breakfast, and Assembly101. Our experiments show that modeling the full context of a video is necessary to obtain the best performance for temporal action segmentation.

Recently, image-to-video (I2V) diffusion models have demonstrated impressive scene understanding and generative quality, incorporating image conditions to guide generation. However, these models primarily animate static images without extending beyond their provided context. Introducing additional constraints, such as camera trajectories, can enhance diversity but often degrades visual quality, limiting their applicability for tasks requiring faithful scene representation. We propose CamContextI2V, an I2V model that integrates multiple image conditions with 3D constraints alongside camera control to enrich both global semantics and fine-grained visual details. This enables more coherent and context-aware video generation. Moreover, we motivate the necessity of temporal awareness for an effective context representation. Our comprehensive study on the RealEstate10K dataset demonstrates improvements in visual quality and camera controllability. We make our code and models publicly available at: https://github.com/LDenninger/CamContextI2V.

Segmenting video content into events provides semantic structures for indexing, retrieval, and summarization. Since motion cues are not available in continuous photo-streams, and annotations in lifelogging are scarce and costly, the frames are usually clustered into events by comparing the visual features between them in an unsupervised way. However, such methodologies are ineffective to deal with heterogeneous events, e.g. taking a walk, and temporary changes in the sight direction, e.g. at a meeting. To address these limitations, we propose Contextual Event Segmentation (CES), a novel segmentation paradigm that uses an LSTM-based generative network to model the photo-stream sequences, predict their visual context, and track their evolution. CES decides whether a frame is an event boundary by comparing the visual context generated from the frames in the past, to the visual context predicted from the future. We implemented CES on a new and massive lifelogging dataset consisting of more than 1.5 million images spanning over 1,723 days. Experiments on the popular EDUB-Seg dataset show that our model outperforms the state-of-the-art by over 16% in f-measure. Furthermore, CES' performance is only 3 points below that of human annotators.

Visual storytelling involves generating a sequence of coherent frames from a textual storyline while maintaining consistency in characters and scenes. Existing autoregressive methods, which rely on previous frame-sentence pairs, struggle with high memory usage, slow generation speeds, and limited context integration. To address these issues, we propose ContextualStory, a novel framework designed to generate coherent story frames and extend frames for visual storytelling. ContextualStory utilizes Spatially-Enhanced Temporal Attention to capture spatial and temporal dependencies, handling significant character movements effectively. Additionally, we introduce a Storyline Contextualizer to enrich context in storyline embedding, and a StoryFlow Adapter to measure scene changes between frames for guiding the model. Extensive experiments on PororoSV and FlintstonesSV datasets demonstrate that ContextualStory significantly outperforms existing SOTA methods in both story visualization and continuation. Code is available at https://github.com/sixiaozheng/ContextualStory.

Creating a vivid video from the event or scenario in our imagination is a truly fascinating experience. Recent advancements in text-to-video synthesis have unveiled the potential to achieve this with prompts only. While text is convenient in conveying the overall scene context, it may be insufficient to control precisely. In this paper, we explore customized video generation by utilizing text as context description and motion structure (e.g. frame-wise depth) as concrete guidance. Our method, dubbed Make-Your-Video, involves joint-conditional video generation using a Latent Diffusion Model that is pre-trained for still image synthesis and then promoted for video generation with the introduction of temporal modules. This two-stage learning scheme not only reduces the computing resources required, but also improves the performance by transferring the rich concepts available in image datasets solely into video generation. Moreover, we use a simple yet effective causal attention mask strategy to enable longer video synthesis, which mitigates the potential quality degradation effectively. Experimental results show the superiority of our method over existing baselines, particularly in terms of temporal coherence and fidelity to users' guidance. In addition, our model enables several intriguing applications that demonstrate potential for practical usage.

The ability to integrate context, including perceptual and temporal cues, plays a pivotal role in grounding the meaning of a linguistic utterance. In order to measure to what extent current vision-and-language models master this ability, we devise a new multimodal challenge, Image Retrieval from Contextual Descriptions (ImageCoDe). In particular, models are tasked with retrieving the correct image from a set of 10 minimally contrastive candidates based on a contextual description. As such, each description contains only the details that help distinguish between images. Because of this, descriptions tend to be complex in terms of syntax and discourse and require drawing pragmatic inferences. Images are sourced from both static pictures and video frames. We benchmark several state-of-the-art models, including both cross-encoders such as ViLBERT and bi-encoders such as CLIP, on ImageCoDe. Our results reveal that these models dramatically lag behind human performance: the best variant achieves an accuracy of 20.9 on video frames and 59.4 on static pictures, compared with 90.8 in humans. Furthermore, we experiment with new model variants that are better equipped to incorporate visual and temporal context into their representations, which achieve modest gains. Our hope is that ImageCoDE will foster progress in grounded language understanding by encouraging models to focus on fine-grained visual differences.

We consider retrieving a specific temporal segment, or moment, from a video given a natural language text description. Methods designed to retrieve whole video clips with natural language determine what occurs in a video but not when. To address this issue, we propose the Moment Context Network (MCN) which effectively localizes natural language queries in videos by integrating local and global video features over time. A key obstacle to training our MCN model is that current video datasets do not include pairs of localized video segments and referring expressions, or text descriptions which uniquely identify a corresponding moment. Therefore, we collect the Distinct Describable Moments (DiDeMo) dataset which consists of over 10,000 unedited, personal videos in diverse visual settings with pairs of localized video segments and referring expressions. We demonstrate that MCN outperforms several baseline methods and believe that our initial results together with the release of DiDeMo will inspire further research on localizing video moments with natural language.

Video Temporal Grounding (VTG) aims to precisely identify video event segments in response to textual queries. The outputs of VTG tasks manifest as sequences of events, each defined by precise timestamps, saliency scores, and textual descriptions. Despite recent advances, a fundamental limitation persists in existing Video Large Language Models (Video-LLMs): they process all task tokens through identical and static pathways, failing to recognize that temporal localization, saliency assessment, and textual generation represent fundamentally distinct tasks requiring specialized processing. To address this, we introduce TimeExpert, a Mixture-of-Experts (MoE)-based Video-LLM that effectively decomposes VTG tasks by dynamically routing task-specific tokens (e.g., timestamps, saliency scores) to specialized experts, with increased computational efficiency. Our design choices enable precise handling of each subtask, leading to improved event modeling across diverse VTG applications. Extensive experiments demonstrate that TimeExpert consistently achieves state-of-the-art performance on various VTG tasks such as Dense Video Captioning, Moment Retrieval, and Video Highlight Detection.

In this paper, we re-examine the task of cross-modal clip-sentence retrieval, where the clip is part of a longer untrimmed video. When the clip is short or visually ambiguous, knowledge of its local temporal context (i.e. surrounding video segments) can be used to improve the retrieval performance. We propose Context Transformer (ConTra); an encoder architecture that models the interaction between a video clip and its local temporal context in order to enhance its embedded representations. Importantly, we supervise the context transformer using contrastive losses in the cross-modal embedding space. We explore context transformers for video and text modalities. Results consistently demonstrate improved performance on three datasets: YouCook2, EPIC-KITCHENS and a clip-sentence version of ActivityNet Captions. Exhaustive ablation studies and context analysis show the efficacy of the proposed method.

Joint audio-video generation models have made substantial progress in visual quality and audio-visual synchronization. However, they still provide limited control over when shot transitions occur and dialogue is spoken. This limitation constrains their application in script-driven content creation, where timing errors can undermine narrative coherence and the viewing experience. Current joint generators align video and audio representations on a shared temporal axis, yet the precise timing of shots and dialogue specified in a structured prompt is encoded only in the prompt's text representation and remains unaligned with the temporal coordinates of either modality. Consequently, video and audio may remain synchronized with each other while both fail to follow the script timeline. This mismatch motivates us to extend temporal alignment beyond video and audio to include the structured script. We therefore introduce Temporal Context Routing (TCR), which maps the script timing onto the shared temporal axis of video and audio generation and routes each prompt's guidance to the corresponding positions in both modalities. Compared with the baseline on 200 test scripts, TCR reduces Shot Boundary MAE by 96%, from 1.11 s to 0.042 s, and raises Dialogue Acc@0.5 s from 28.3% to 84.1%. TCR achieves these improvements while maintaining visual quality and audio-visual synchronization comparable to those of the baselines. A user study further shows that participants prefer TCR on all five evaluated dimensions.

Building on the advances of language models, Large Multimodal Models (LMMs) have contributed significant improvements in video understanding. While the current video LMMs utilize advanced Large Language Models (LLMs), they rely on either image or video encoders to process visual inputs, each of which has its own limitations. Image encoders excel at capturing rich spatial details from frame sequences but lack explicit temporal context, which can be important in videos with intricate action sequences. On the other hand, video encoders provide temporal context but are often limited by computational constraints that lead to processing only sparse frames at lower resolutions, resulting in reduced contextual and spatial understanding. To this end, we introduce VideoGPT+, which combines the complementary benefits of the image encoder (for detailed spatial understanding) and the video encoder (for global temporal context modeling). The model processes videos by dividing them into smaller segments and applies an adaptive pooling strategy on features extracted by both image and video encoders. Our architecture showcases improved performance across multiple video benchmarks, including VCGBench, MVBench and Zero-shot question-answering. Further, we develop 112K video-instruction set using a novel semi-automatic annotation pipeline which further improves the model performance. Additionally, to comprehensively evaluate video LMMs, we present VCGBench-Diverse, covering 18 broad video categories such as lifestyle, sports, science, gaming, and surveillance videos. This benchmark with 4,354 question-answer pairs evaluates the generalization of existing LMMs on dense video captioning, spatial and temporal understanding, and complex reasoning, ensuring comprehensive assessment across diverse video types and dynamics. Code: https://github.com/mbzuai-oryx/VideoGPT-plus.

In static monitoring cameras, useful contextual information can stretch far beyond the few seconds typical video understanding models might see: subjects may exhibit similar behavior over multiple days, and background objects remain static. Due to power and storage constraints, sampling frequencies are low, often no faster than one frame per second, and sometimes are irregular due to the use of a motion trigger. In order to perform well in this setting, models must be robust to irregular sampling rates. In this paper we propose a method that leverages temporal context from the unlabeled frames of a novel camera to improve performance at that camera. Specifically, we propose an attention-based approach that allows our model, Context R-CNN, to index into a long term memory bank constructed on a per-camera basis and aggregate contextual features from other frames to boost object detection performance on the current frame. We apply Context R-CNN to two settings: (1) species detection using camera traps, and (2) vehicle detection in traffic cameras, showing in both settings that Context R-CNN leads to performance gains over strong baselines. Moreover, we show that increasing the contextual time horizon leads to improved results. When applied to camera trap data from the Snapshot Serengeti dataset, Context R-CNN with context from up to a month of images outperforms a single-frame baseline by 17.9% mAP, and outperforms S3D (a 3d convolution based baseline) by 11.2% mAP.

Neural Video Compression has emerged in recent years, with condition-based frameworks outperforming traditional codecs. However, most existing methods rely solely on the previous frame's features to predict temporal context, leading to two critical issues. First, the short reference window misses long-term dependencies and fine texture details. Second, propagating only feature-level information accumulates errors over frames, causing prediction inaccuracies and loss of subtle textures. To address these, we propose the Long-term Spatio-Temporal Enhanced Context (L-STEC) method. We first extend the reference chain with LSTM to capture long-term dependencies. We then incorporate warped spatial context from the pixel domain, fusing spatio-temporal information through a multi-receptive field network to better preserve reference details. Experimental results show that L-STEC significantly improves compression by enriching contextual information, achieving 37.01% bitrate savings in PSNR and 31.65% in MS-SSIM compared to DCVC-TCM, outperforming both VTM-17.0 and DCVC-FM and establishing new state-of-the-art performance.

Video temporal grounding aims to pinpoint a video segment that matches the query description. Despite the recent advance in short-form videos (e.g., in minutes), temporal grounding in long videos (e.g., in hours) is still at its early stage. To address this challenge, a common practice is to employ a sliding window, yet can be inefficient and inflexible due to the limited number of frames within the window. In this work, we propose an end-to-end framework for fast temporal grounding, which is able to model an hours-long video with one-time network execution. Our pipeline is formulated in a coarse-to-fine manner, where we first extract context knowledge from non-overlapped video clips (i.e., anchors), and then supplement the anchors that highly response to the query with detailed content knowledge. Besides the remarkably high pipeline efficiency, another advantage of our approach is the capability of capturing long-range temporal correlation, thanks to modeling the entire video as a whole, and hence facilitates more accurate grounding. Experimental results suggest that, on the long-form video datasets MAD and Ego4d, our method significantly outperforms state-of-the-arts, and achieves 14.6times / 102.8times higher efficiency respectively. Project can be found at https://github.com/afcedf/SOONet.git.

Understanding time is crucial for understanding events expressed in natural language. Because people rarely say the obvious, it is often necessary to have commonsense knowledge about various temporal aspects of events, such as duration, frequency, and temporal order. However, this important problem has so far received limited attention. This paper systematically studies this temporal commonsense problem. Specifically, we define five classes of temporal commonsense, and use crowdsourcing to develop a new dataset, MCTACO, that serves as a test set for this task. We find that the best current methods used on MCTACO are still far behind human performance, by about 20%, and discuss several directions for improvement. We hope that the new dataset and our study here can foster more future research on this topic.

Large-scale vision-language models (VLMs) such as CLIP have gained popularity for their generalizable and expressive multimodal representations. By leveraging large-scale training data with diverse textual metadata, VLMs acquire open-vocabulary capabilities, solving tasks beyond their training scope. This paper investigates the temporal awareness of VLMs, assessing their ability to position visual content in time. We introduce TIME10k, a benchmark dataset of over 10,000 images with temporal ground truth, and evaluate the time-awareness of 37 VLMs by a novel methodology. Our investigation reveals that temporal information is structured along a low-dimensional, non-linear manifold in the VLM embedding space. Based on this insight, we propose methods to derive an explicit ``timeline'' representation from the embedding space. These representations model time and its chronological progression and thereby facilitate temporal reasoning tasks. Our timeline approaches achieve competitive to superior accuracy compared to a prompt-based baseline while being computationally efficient. All code and data are available at https://tekayanidham.github.io/timeline-page/.

Video Temporal Grounding (VTG) aims to extract relevant video segments based on a given natural language query. Recently, zero-shot VTG methods have gained attention by leveraging pretrained vision-language models (VLMs) to localize target moments without additional training. However, existing approaches suffer from semantic fragmentation, where temporally continuous frames sharing the same semantics are split across multiple segments. When segments are fragmented, it becomes difficult to predict an accurate target moment that aligns with the text query. Also, they rely on skewed similarity distributions for localization, making it difficult to select the optimal segment. Furthermore, they heavily depend on the use of LLMs which require expensive inferences. To address these limitations, we propose a TAG, a simple yet effective Temporal-Aware approach for zero-shot video temporal Grounding, which incorporates temporal pooling, temporal coherence clustering, and similarity adjustment. Our proposed method effectively captures the temporal context of videos and addresses distorted similarity distributions without training. Our approach achieves state-of-the-art results on Charades-STA and ActivityNet Captions benchmark datasets without rely on LLMs. Our code is available at https://github.com/Nuetee/TAG

The recent success of Large Language Models (LLMs) has prompted the extension to the multimodal domain, developing image-text Multimodal LLMs (MLLMs) and then video-text models. In this work, we investigate the challenge of contextual and temporal comprehension in video-language models by exploring the task of temporal localization in videos. To address this problem, prior works have developed complex task-specific architectures, novel modules to embed time into MLLMs, or leveraged additional input signals such as video transcripts to best encode contextual and temporal information. We find that most of these efforts are surpassed by a much simpler design. We introduce Chrono, a universal sequence blueprint that can be applied to any image-text pretrained MLLM. In extensive experiments spanning different MLLM architectures and sizes, finetuning and zero-shot settings, we demonstrate new state-of-the-art results in moment retrieval on the widely used benchmarks Charades-STA, QVHighlights, and ActivityNet Captions, as well as in grounded video question answering on NExT-GQA.

Understanding temporal dynamics is critical for conversational agents, enabling effective content analysis and informed decision-making. However, time-aware datasets, particularly for persona-grounded conversations, are still limited, which narrows their scope and diminishes their complexity. To address this gap, we introduce MTPChat, a multimodal, time-aware persona dialogue dataset that integrates linguistic, visual, and temporal elements within dialogue and persona memory. Leveraging MTPChat, we propose two time-sensitive tasks: Temporal Next Response Prediction (TNRP) and Temporal Grounding Memory Prediction (TGMP), both designed to assess a model's ability to understand implicit temporal cues and dynamic interactions. Additionally, we present an innovative framework featuring an adaptive temporal module to effectively integrate multimodal streams and capture temporal dependencies. Experimental results validate the challenges posed by MTPChat and demonstrate the effectiveness of our framework in multimodal time-sensitive scenarios.

Recent advancements in time series forecasting have explored augmenting models with text or vision modalities to improve accuracy. While text provides contextual understanding, it often lacks fine-grained temporal details. Conversely, vision captures intricate temporal patterns but lacks semantic context, limiting the complementary potential of these modalities. To address this, we propose \method, a novel multimodal framework that leverages pre-trained Vision-Language Models (VLMs) to bridge temporal, visual, and textual modalities for enhanced forecasting. Our framework comprises three key components: (1) a Retrieval-Augmented Learner, which extracts enriched temporal features through memory bank interactions; (2) a Vision-Augmented Learner, which encodes time series as informative images; and (3) a Text-Augmented Learner, which generates contextual textual descriptions. These components collaborate with frozen pre-trained VLMs to produce multimodal embeddings, which are then fused with temporal features for final prediction. Extensive experiments demonstrate that Time-VLM achieves superior performance, particularly in few-shot and zero-shot scenarios, thereby establishing a new direction for multimodal time series forecasting. Code is available at https://github.com/CityMind-Lab/ICML25-TimeVLM.

Temporal modeling is crucial for various video learning tasks. Most recent approaches employ either factorized (2D+1D) or joint (3D) spatial-temporal operations to extract temporal contexts from the input frames. While the former is more efficient in computation, the latter often obtains better performance. In this paper, we attribute this to a dilemma between the sufficiency and the efficiency of interactions among various positions in different frames. These interactions affect the extraction of task-relevant information shared among frames. To resolve this issue, we prove that frame-by-frame alignments have the potential to increase the mutual information between frame representations, thereby including more task-relevant information to boost effectiveness. Then we propose Alignment-guided Temporal Attention (ATA) to extend 1-dimensional temporal attention with parameter-free patch-level alignments between neighboring frames. It can act as a general plug-in for image backbones to conduct the action recognition task without any model-specific design. Extensive experiments on multiple benchmarks demonstrate the superiority and generality of our module.

Temporal Grounding (TG) aims to localize video segments corresponding to a textual query. Prior research predominantly focuses on single-segment retrieval. Real-world scenarios, however, often require localizing multiple disjoint segments for a single query -- a setting we term One-to-Many Temporal Grounding (OMTG). Previous state-of-the-art MLLMs, optimized for one-to-one settings, struggle in this context, often yielding near-zero scores due to a lack of event cardinality perception. To bridge this gap, we present a systematic solution with three key contributions. First, we establish the first comprehensive OMTG benchmark, introducing Count Accuracy (C-Acc) and Effective Temporal F1 (EtF1) as evaluation metrics. Second, we curate a high-quality OMTG dataset comprising 56k samples through a sophisticated construction pipeline. Third, we develop novel temporal and caption reward functions specifically designed for OMTG. In particular, the caption reward leverages Chain-of-Thought reasoning over dense video captions to explicitly guide policy optimization toward both preciseness and completeness. Extensive experiments show our model achieves a new state-of-the-art EtF1 of 43.65\% on OMTG Bench, outperforming Gemini 2.5 Pro and Seed-1.8 by 15.85\% and 15.61\%, respectively.

World simulation has gained increasing popularity due to its ability to model virtual environments and predict the consequences of actions. However, the limited temporal context window often leads to failures in maintaining long-term consistency, particularly in preserving 3D spatial consistency. In this work, we present WorldMem, a framework that enhances scene generation with a memory bank consisting of memory units that store memory frames and states (e.g., poses and timestamps). By employing a memory attention mechanism that effectively extracts relevant information from these memory frames based on their states, our method is capable of accurately reconstructing previously observed scenes, even under significant viewpoint or temporal gaps. Furthermore, by incorporating timestamps into the states, our framework not only models a static world but also captures its dynamic evolution over time, enabling both perception and interaction within the simulated world. Extensive experiments in both virtual and real scenarios validate the effectiveness of our approach.

As humans, we understand events in the visual world contextually, performing multimodal reasoning across time to make inferences about the past, present, and future. We introduce MERLOT, a model that learns multimodal script knowledge by watching millions of YouTube videos with transcribed speech -- in an entirely label-free, self-supervised manner. By pretraining with a mix of both frame-level (spatial) and video-level (temporal) objectives, our model not only learns to match images to temporally corresponding words, but also to contextualize what is happening globally over time. As a result, MERLOT exhibits strong out-of-the-box representations of temporal commonsense, and achieves state-of-the-art performance on 12 different video QA datasets when finetuned. It also transfers well to the world of static images, allowing models to reason about the dynamic context behind visual scenes. On Visual Commonsense Reasoning, MERLOT answers questions correctly with 80.6% accuracy, outperforming state-of-the-art models of similar size by over 3%, even those that make heavy use of auxiliary supervised data (like object bounding boxes). Ablation analyses demonstrate the complementary importance of: 1) training on videos versus static images; 2) scaling the magnitude and diversity of the pretraining video corpus; and 3) using diverse objectives that encourage full-stack multimodal reasoning, from the recognition to cognition level.

Time awareness is a fundamental ability of omni large language models, especially for understanding long videos and answering complex questions. Previous approaches mainly target vision-language scenarios and focus on the explicit temporal grounding questions, such as identifying when a visual event occurs or determining what event happens at aspecific time. However, they often make insufficient use of the audio modality, and overlook implicit temporal grounding across modalities--for example, identifying what is visually present when a character speaks, or determining what is said when a visual event occurs--despite such cross-modal temporal relations being prevalent in real-world scenarios. In this paper, we propose ChronusOmni, an omni large language model designed to enhance temporal awareness for both explicit and implicit audiovisual temporal grounding. First, we interleave text-based timestamp tokens with visual and audio representations at each time unit, enabling unified temporal modeling across modalities. Second, to enforce correct temporal ordering and strengthen fine-grained temporal reasoning, we incorporate reinforcement learning with specially designed reward functions. Moreover, we construct ChronusAV, a temporally-accurate, modality-complete, and cross-modal-aligned dataset to support the training and evaluation on audiovisual temporal grounding task. Experimental results demonstrate that ChronusOmni achieves state-of-the-art performance on ChronusAV with more than 30% improvement and top results on most metrics upon other temporal grounding benchmarks. This highlights the strong temporal awareness of our model across modalities, while preserving general video and audio understanding capabilities.

This analysis explores the temporal sequencing of objects in a movie trailer. Temporal sequencing of objects in a movie trailer (e.g., a long shot of an object vs intermittent short shots) can convey information about the type of movie, plot of the movie, role of the main characters, and the filmmakers cinematographic choices. When combined with historical customer data, sequencing analysis can be used to improve predictions of customer behavior. E.g., a customer buys tickets to a new movie and maybe the customer has seen movies in the past that contained similar sequences. To explore object sequencing in movie trailers, we propose a video convolutional network to capture actions and scenes that are predictive of customers' preferences. The model learns the specific nature of sequences for different types of objects (e.g., cars vs faces), and the role of sequences in predicting customer future behavior. We show how such a temporal-aware model outperforms simple feature pooling methods proposed in our previous works and, importantly, demonstrate the additional model explain-ability allowed by such a model.

Natural scenes provide us with rich contexts for object recognition and reference. In particular, knowing what type of scene one is looking at generates expectations about which objects will occur, and what their spatial configuration should be. Do Vision-Language Models (VLMs) learn to rely on scene contexts in a similar way, when generating references to objects? To address this question, we introduce the Common Objects Out-of-Context (COOCO) dataset and test to what extent VLMs rely on scene context to refer to objects under different degrees of scene-object congruency, and different perturbations. Our findings show that models leverage scene context adaptively, depending on both the semantic relatedness between object and scene and the level of noise. In particular, models rely more on context under high target-scene congruence or when objects are degraded. Attention analysis reveals that successful object categorisation involves increased focus on the target in mid-level layers, especially under moderate noise, suggesting that VLMs dynamically balance local and contextual information for reference generation. We make our dataset, code and models available at https://github.com/cs-nlp-uu/scenereg{https://github.com/cs-nlp-uu/scenereg}.

Cinematic video production requires control over scene-subject composition and camera movement, but live-action shooting remains costly due to the need for constructing physical sets. To address this, we introduce the task of cinematic video generation with decoupled scene context: given multiple images of a static environment, the goal is to synthesize high-quality videos featuring dynamic subject while preserving the underlying scene consistency and following a user-specified camera trajectory. We present CineScene, a framework that leverages implicit 3D-aware scene representation for cinematic video generation. Our key innovation is a novel context conditioning mechanism that injects 3D-aware features in an implicit way: By encoding scene images into visual representations through VGGT, CineScene injects spatial priors into a pretrained text-to-video generation model by additional context concatenation, enabling camera-controlled video synthesis with consistent scenes and dynamic subjects. To further enhance the model's robustness, we introduce a simple yet effective random-shuffling strategy for the input scene images during training. To address the lack of training data, we construct a scene-decoupled dataset with Unreal Engine 5, containing paired videos of scenes with and without dynamic subjects, panoramic images representing the underlying static scene, along with their camera trajectories. Experiments show that CineScene achieves state-of-the-art performance in scene-consistent cinematic video generation, handling large camera movements and demonstrating generalization across diverse environments.

Geo-temporal understanding, the ability to infer location, time, and contextual properties from visual input alone, underpins applications such as disaster management, traffic planning, embodied navigation, world modeling, and geography education. Although recent vision-language models (VLMs) have advanced image geo-localization using cues like landmarks and road signs, their ability to reason about temporal signals and physically grounded spatial cues remains limited. To address this gap, we introduce TimeSpot, a benchmark for evaluating real-world geo-temporal reasoning in VLMs. TimeSpot comprises 1,455 ground-level images from 80 countries and requires structured prediction of temporal attributes (season, month, time of day, daylight phase) and geographic attributes (continent, country, climate zone, environment type, latitude-longitude) directly from visual evidence. It also includes spatial-temporal reasoning tasks that test physical plausibility under real-world uncertainty. Evaluations of state-of-the-art open- and closed-source VLMs show low performance, particularly for temporal inference. While supervised fine-tuning yields improvements, results remain insufficient, highlighting the need for new methods to achieve robust, physically grounded geo-temporal understanding. TimeSpot is available at: https://TimeSpot-GT.github.io.

Objective: To enable context-aware computer assistance in the operating room of the future, cognitive systems need to understand automatically which surgical phase is being performed by the medical team. The primary source of information for surgical phase recognition is typically video, which presents two challenges: extracting meaningful features from the video stream and effectively modeling temporal information in the sequence of visual features. Methods: For temporal modeling, attention mechanisms have gained popularity due to their ability to capture long-range dependencies. In this paper, we explore design choices for attention in existing temporal models for surgical phase recognition and propose a novel approach that uses attention more effectively and does not require hand-crafted constraints: TUNeS, an efficient and simple temporal model that incorporates self-attention at the core of a convolutional U-Net structure. In addition, we propose to train the feature extractor, a standard CNN, together with an LSTM on preferably long video segments, i.e., with long temporal context. Results: In our experiments, almost all temporal models performed better on top of feature extractors that were trained with longer temporal context. On these contextualized features, TUNeS achieves state-of-the-art results on the Cholec80 dataset. Conclusion: This study offers new insights on how to use attention mechanisms to build accurate and efficient temporal models for surgical phase recognition. Significance: Implementing automatic surgical phase recognition is essential to automate the analysis and optimization of surgical workflows and to enable context-aware computer assistance during surgery, thus ultimately improving patient care.

In the evolving field of Natural Language Processing, understanding the temporal context of text is increasingly crucial. This study investigates methods to incorporate temporal information during pre-training, aiming to achieve effective time-aware language representation for improved performance on time-related tasks. In contrast to common pre-trained models like BERT, which rely on synchronic document collections such as BookCorpus and Wikipedia, our research introduces BiTimeBERT 2.0, a novel language model pre-trained on a temporal news article collection. BiTimeBERT 2.0 utilizes this temporal news collection, focusing on three innovative pre-training objectives: Time-Aware Masked Language Modeling (TAMLM), Document Dating (DD), and Time-Sensitive Entity Replacement (TSER). Each objective targets a unique aspect of temporal information. TAMLM is designed to enhance the understanding of temporal contexts and relations, DD integrates document timestamps as chronological markers, and TSER focuses on the temporal dynamics of "Person" entities, recognizing their inherent temporal significance. The experimental results consistently demonstrate that BiTimeBERT 2.0 outperforms models like BERT and other existing pre-trained models, achieving substantial gains across a variety of downstream NLP tasks and applications where time plays a pivotal role.

Text to video generation has emerged as a critical frontier in generative artificial intelligence, yet existing approaches struggle with maintaining temporal consistency, compositional understanding, and fine grained control over visual narratives. We present MOVAI (Multimodal Original Video AI), a novel hierarchical framework that integrates compositional scene understanding with temporal aware diffusion models for high fidelity text to video synthesis. Our approach introduces three key innovations: (1) a Compositional Scene Parser (CSP) that decomposes textual descriptions into hierarchical scene graphs with temporal annotations, (2) a Temporal-Spatial Attention Mechanism (TSAM) that ensures coherent motion dynamics across frames while preserving spatial details, and (3) a Progressive Video Refinement (PVR) module that iteratively enhances video quality through multi-scale temporal reasoning. Extensive experiments on standard benchmarks demonstrate that MOVAI achieves state-of-the-art performance, improving video quality metrics by 15.3% in LPIPS, 12.7% in FVD, and 18.9% in user preference studies compared to existing methods. Our framework shows particular strength in generating complex multi-object scenes with realistic temporal dynamics and fine-grained semantic control.

We study object interaction anticipation in egocentric videos. This task requires an understanding of the spatio-temporal context formed by past actions on objects, coined action context. We propose TransFusion, a multimodal transformer-based architecture. It exploits the representational power of language by summarizing the action context. TransFusion leverages pre-trained image captioning and vision-language models to extract the action context from past video frames. This action context together with the next video frame is processed by the multimodal fusion module to forecast the next object interaction. Our model enables more efficient end-to-end learning. The large pre-trained language models add common sense and a generalisation capability. Experiments on Ego4D and EPIC-KITCHENS-100 show the effectiveness of our multimodal fusion model. They also highlight the benefits of using language-based context summaries in a task where vision seems to suffice. Our method outperforms state-of-the-art approaches by 40.4% in relative terms in overall mAP on the Ego4D test set. We validate the effectiveness of TransFusion via experiments on EPIC-KITCHENS-100. Video and code are available at https://eth-ait.github.io/transfusion-proj/.

Video World Models are interactive video generation models that predict future world states based on user actions and history video frames. A critical challenge in video world models is the lack of memory, causing inconsistent generated scenes over extended durations. Previous methods explored rule-based context frame retrieval as memory, but they fail to generalize in scenarios with scene occlusions and dynamic objects. We propose MemLearner, a learning-based adaptive context query method using query tokens to bridge context and predicted tokens. By leveraging the video generation model itself for context querying, MemLearner exploits pre-trained visual priors without training additional modules from scratch, and incorporates efficient strategies for training and inference. We collect a dataset of long videos with scene occlusions and dynamic objects, paired with camera pose annotations, and propose a multi-dataset training strategy leveraging both annotated rendered and unannotated real-world videos. Extensive experiments demonstrate that MemLearner significantly outperforms prior video world models in terms of scene consistency and memory, particularly under challenging occlusion and dynamic scenarios.

Spatio-temporal grounding describes the task of localizing events in space and time, e.g., in video data, based on verbal descriptions only. Models for this task are usually trained with human-annotated sentences and bounding box supervision. This work addresses this task from a multimodal supervision perspective, proposing a framework for spatio-temporal action grounding trained on loose video and subtitle supervision only, without human annotation. To this end, we combine local representation learning, which focuses on leveraging fine-grained spatial information, with a global representation encoding that captures higher-level representations and incorporates both in a joint approach. To evaluate this challenging task in a real-life setting, a new benchmark dataset is proposed providing dense spatio-temporal grounding annotations in long, untrimmed, multi-action instructional videos for over 5K events. We evaluate the proposed approach and other methods on the proposed and standard downstream tasks showing that our method improves over current baselines in various settings, including spatial, temporal, and untrimmed multi-action spatio-temporal grounding.

Long video understanding remains a fundamental challenge for multimodal large language models (MLLMs), particularly in tasks requiring precise temporal reasoning and event localization. Existing approaches typically adopt uniform frame sampling and rely on implicit position encodings to model temporal order. However, these methods struggle with long-range dependencies, leading to critical information loss and degraded temporal comprehension. In this paper, we propose Dynamic Absolute Time Enhancement (DATE) that enhances temporal awareness in MLLMs through the Timestamp Injection Mechanism (TIM) and a semantically guided Temporal-Aware Similarity Sampling (TASS) strategy. Specifically, we interleave video frame embeddings with textual timestamp tokens to construct a continuous temporal reference system. We further reformulate the video sampling problem as a vision-language retrieval task and introduce a two-stage algorithm to ensure both semantic relevance and temporal coverage: enriching each query into a descriptive caption to better align with the vision feature, and sampling key event with a similarity-driven temporally regularized greedy strategy. Our method achieves remarkable improvements w.r.t. absolute time understanding and key event localization, resulting in state-of-the-art performance among 7B and 72B models on hour-long video benchmarks. Particularly, our 7B model even exceeds many 72B models on some benchmarks.

Understanding videos requires more than answering open ended questions, it demands the ability to pinpoint when events occur and how entities interact across time. While recent Video LLMs have achieved remarkable progress in holistic reasoning, they remain coarse in temporal perception: timestamps are encoded only implicitly, frame level features are weak in capturing continuity, and language vision alignment often drifts from the entities of interest. In this paper, we present Grounded VideoDiT, a Video LLM designed to overcome these limitations by introducing three key innovations. First, a Diffusion Temporal Latent (DTL) encoder enhances boundary sensitivity and maintains temporal consistency. Second, object grounded representations explicitly bind query entities to localized visual evidence, strengthening alignment. Third, a mixed token scheme with discrete temporal tokens provides explicit timestamp modeling, enabling fine grained temporal reasoning. Together, these designs equip Grounded VideoDiT with robust grounding capabilities, as validated by state of the art results on Charades STA, NExT GQA, and multiple VideoQA benchmarks.

Temporal information has long been considered to be essential for perception. While there is extensive research on the role of image information for perceptual tasks, the role of the temporal dimension remains less well understood: What can we learn about the world from long-term motion information? What properties does long-term motion information have for visual learning? We leverage recent success in point-track estimation, which offers an excellent opportunity to learn temporal representations and experiment on a variety of perceptual tasks. We draw 3 clear lessons: 1) Long-term motion representations contain information to understand actions, but also objects, materials, and spatial information, often even better than images. 2) Long-term motion representations generalize far better than image representations in low-data settings and in zero-shot tasks. 3) The very low dimensionality of motion information makes motion representations a better trade-off between GFLOPs and accuracy than standard video representations, and used together they achieve higher performance than video representations alone. We hope these insights will pave the way for the design of future models that leverage the power of long-term motion information for perception.

Pre-training on large-scale video data has become a common recipe for learning transferable spatiotemporal representations in recent years. Despite some progress, existing methods are mostly limited to highly curated datasets (e.g., K400) and exhibit unsatisfactory out-of-the-box representations. We argue that it is due to the fact that they only capture pixel-level knowledge rather than spatiotemporal semantics, which hinders further progress in video understanding. Inspired by the great success of image-text pre-training (e.g., CLIP), we take the first step to exploit language semantics to boost transferable spatiotemporal representation learning. We introduce a new pretext task, Turning to Video for Transcript Sorting (TVTS), which sorts shuffled ASR scripts by attending to learned video representations. We do not rely on descriptive captions and learn purely from video, i.e., leveraging the natural transcribed speech knowledge to provide noisy but useful semantics over time. Our method enforces the vision model to contextualize what is happening over time so that it can re-organize the narrative transcripts, and can seamlessly apply to large-scale uncurated video data in the real world. Our method demonstrates strong out-of-the-box spatiotemporal representations on diverse benchmarks, e.g., +13.6% gains over VideoMAE on SSV2 via linear probing. The code is available at https://github.com/TencentARC/TVTS.

We investigate a challenging task of dynamic scene geometry estimation, which requires representing both spatial and temporal features. Typically, existing methods align the two features into a unified latent space to model scene geometry. However, this unified paradigm suffers from potential mismatched representation due to the heterogeneous nature between spatial and temporal features. In this work, we propose 4D-VGGT, a general foundation model with divide-and-conquer spatiotemporal representation for dynamic scene geometry. Our model is divided into three aspects: 1) Multi-setting input. We design an adaptive visual grid that supports input sequences with arbitrary numbers of views and time steps. 2) Multi-level representation. We propose a cross-view global fusion for spatial representation and a cross-time local fusion for temporal representation. 3) Multi-task prediction. We append multiple task-specific heads to spatiotemporal representations, enabling a comprehensive visual geometry estimation for dynamic scenes. Under this unified framework, these components enhance the feature discriminability and application universality of our model for dynamic scenes. In addition, we integrate multiple geometry datasets to train our model and conduct extensive experiments to verify the effectiveness of our method across various tasks on multiple dynamic scene geometry benchmarks.

Text-to-video editing aims to edit the visual appearance of a source video conditional on textual prompts. A major challenge in this task is to ensure that all frames in the edited video are visually consistent. Most recent works apply advanced text-to-image diffusion models to this task by inflating 2D spatial attention in the U-Net into spatio-temporal attention. Although temporal context can be added through spatio-temporal attention, it may introduce some irrelevant information for each patch and therefore cause inconsistency in the edited video. In this paper, for the first time, we introduce optical flow into the attention module in the diffusion model's U-Net to address the inconsistency issue for text-to-video editing. Our method, FLATTEN, enforces the patches on the same flow path across different frames to attend to each other in the attention module, thus improving the visual consistency in the edited videos. Additionally, our method is training-free and can be seamlessly integrated into any diffusion-based text-to-video editing methods and improve their visual consistency. Experiment results on existing text-to-video editing benchmarks show that our proposed method achieves the new state-of-the-art performance. In particular, our method excels in maintaining the visual consistency in the edited videos.

Despite progress in video-language modeling, the computational challenge of interpreting long-form videos in response to task-specific linguistic queries persists, largely due to the complexity of high-dimensional video data and the misalignment between language and visual cues over space and time. To tackle this issue, we introduce a novel approach called Language-guided Spatial-Temporal Prompt Learning (LSTP). This approach features two key components: a Temporal Prompt Sampler (TPS) with optical flow prior that leverages temporal information to efficiently extract relevant video content, and a Spatial Prompt Solver (SPS) that adeptly captures the intricate spatial relationships between visual and textual elements. By harmonizing TPS and SPS with a cohesive training strategy, our framework significantly enhances computational efficiency, temporal understanding, and spatial-temporal alignment. Empirical evaluations across two challenging tasks--video question answering and temporal question grounding in videos--using a variety of video-language pretrainings (VLPs) and large language models (LLMs) demonstrate the superior performance, speed, and versatility of our proposed LSTP paradigm.

Video temporal grounding (VTG) aims to localize the continuous video interval described by a natural-language query. However, current VLM-based methods typically produce this interval indirectly through two endpoint outputs, represented either as discrete timestamp tokens or continuous boundary coordinates. These formulations differ in how endpoints are encoded, but not in what is predicted: the event interval remains a derived object, while interval validity, duration, and interval-level similarity are handled only implicitly. We propose TimePLE, which reformulates VTG from endpoint prediction to interval-native grounding by predicting a single joint distribution over valid temporal intervals. TimePLE maps each interval to a point in a canonical position-duration square, where every support point corresponds to a valid span and neighboring points represent geometrically similar intervals. Given a video and query, the VLM generates a single latent <|TIMESPAN|> token whose hidden state is decoded into a joint interval distribution, refined through duration-aware coordinate correction, and converted into continuous boundaries. The same interval representation is used to encode input temporal anchors, aligning video-side temporal evidence with output-side span prediction. To reliably align the latent span representation with complete event intervals, we curate 90K-scale grounded samples and human-verify 3K-scale benchmark annotations. Experiments across four VTG benchmarks show that TimePLE consistently outperforms endpoint prediction baselines, achieving an average mIoU of 58.9, with clear gains on short-duration and medium-duration events.

Temporal reasoning is a critical challenge in video-language understanding, as it requires models to align semantic concepts consistently across time. While existing large vision-language models (LVLMs) and large language models (LLMs) excel at static tasks, they struggle to capture dynamic interactions and temporal dependencies in video sequences. In this work, we propose Temporal Semantic Alignment via Dynamic Prompting (TSADP), a novel framework that enhances temporal reasoning capabilities through dynamic task-specific prompts and temporal contrastive learning. TSADP leverages a Dynamic Prompt Generator (DPG) to encode fine-grained temporal relationships and a Temporal Contrastive Loss (TCL) to align visual and textual embeddings across time. We evaluate our method on the VidSitu dataset, augmented with enriched temporal annotations, and demonstrate significant improvements over state-of-the-art models in tasks such as Intra-Video Entity Association, Temporal Relationship Understanding, and Chronology Prediction. Human evaluations further confirm TSADP's ability to generate coherent and semantically accurate descriptions. Our analysis highlights the robustness, efficiency, and practical utility of TSADP, making it a step forward in the field of video-language understanding.

Multimodal Large Language Models (MLLMs) have demonstrated significant capabilities in image understanding, but long-video are constrained by context windows and computational cost. Uniform frame sampling often leads to substantial information loss. Meanwhile existing keyframe selection methods such as text-frame retrieval or RL-based frame optimization typically yield sparse and temporally disjointed frames, overlooking scene continuity and lacking flexibility for multi-scale frame selection. To address these limitations, we introduce K-frames, a novel paradigm for scene-driven keyframe selection that preserves temporal continuity. Instead of selecting individual frames, K-frames predicts semantically coherent, query-relevant clips, which enables any-k keyframes selection to meet diverse user budgets. To achieve this approach, we first introduce PeakClips, a dataset of 200K video highlights conditioned by query. Building on this dataset, K-frames learns clip2frame selection using a three-stage progressive curriculum. It involves two Supervised Fine-Tuning stages for temporal grounding and key-clip perception, followed by a Reinforcement Learning stage that directly optimizes the scene-driven prediction policy for downstream task without further annotations. Extensive experiments on major long-video understanding benchmarks demonstrate that K-frames provides an effective, interpretable, and plug-and-play solution for keyframe selection at various scales. Our dataset and model will be available.

Recent advances in interactive video generation have shown promising results, yet existing approaches struggle with scene-consistent memory capabilities in long video generation due to limited use of historical context. In this work, we propose Context-as-Memory, which utilizes historical context as memory for video generation. It includes two simple yet effective designs: (1) storing context in frame format without additional post-processing; (2) conditioning by concatenating context and frames to be predicted along the frame dimension at the input, requiring no external control modules. Furthermore, considering the enormous computational overhead of incorporating all historical context, we propose the Memory Retrieval module to select truly relevant context frames by determining FOV (Field of View) overlap between camera poses, which significantly reduces the number of candidate frames without substantial information loss. Experiments demonstrate that Context-as-Memory achieves superior memory capabilities in interactive long video generation compared to SOTAs, even generalizing effectively to open-domain scenarios not seen during training. The link of our project page is https://context-as-memory.github.io/.

Vision-language tracking aims to locate the target object in the video sequence using a template patch and a language description provided in the initial frame. To achieve robust tracking, especially in complex long-term scenarios that reflect real-world conditions as recently highlighted by MGIT, it is essential not only to characterize the target features but also to utilize the context features related to the target. However, the visual and textual target-context cues derived from the initial prompts generally align only with the initial target state. Due to their dynamic nature, target states are constantly changing, particularly in complex long-term sequences. It is intractable for these cues to continuously guide Vision-Language Trackers (VLTs). Furthermore, for the text prompts with diverse expressions, our experiments reveal that existing VLTs struggle to discern which words pertain to the target or the context, complicating the utilization of textual cues. In this work, we present a novel tracker named ATCTrack, which can obtain multimodal cues Aligned with the dynamic target states through comprehensive Target-Context feature modeling, thereby achieving robust tracking. Specifically, (1) for the visual modality, we propose an effective temporal visual target-context modeling approach that provides the tracker with timely visual cues. (2) For the textual modality, we achieve precise target words identification solely based on textual content, and design an innovative context words calibration method to adaptively utilize auxiliary context words. (3) We conduct extensive experiments on mainstream benchmarks and ATCTrack achieves a new SOTA performance. The code and models will be released at: https://github.com/XiaokunFeng/ATCTrack.

Large video-language models (LVLMs) have shown remarkable performance across various video-language tasks. However, they encounter significant challenges when processing long videos because of the large number of video frames involved. Downsampling long videos in either space or time can lead to visual hallucinations, making it difficult to accurately interpret long videos. Motivated by human hierarchical temporal search strategies, we propose TimeSearch, a novel framework enabling LVLMs to understand long videos in a human-like manner. TimeSearch integrates two human-like primitives into a unified autoregressive LVLM: 1) Spotlight efficiently identifies relevant temporal events through a Temporal-Augmented Frame Representation (TAFR), explicitly binding visual features with timestamps; 2) Reflection evaluates the correctness of the identified events, leveraging the inherent temporal self-reflection capabilities of LVLMs. TimeSearch progressively explores key events and prioritizes temporal search based on reflection confidence. Extensive experiments on challenging long-video benchmarks confirm that TimeSearch substantially surpasses previous state-of-the-art, improving the accuracy from 41.8\% to 51.5\% on the LVBench. Additionally, experiments on temporal grounding demonstrate that appropriate TAFR is adequate to effectively stimulate the surprising temporal grounding ability of LVLMs in a simpler yet versatile manner, which improves mIoU on Charades-STA by 11.8\%. The code will be released.

Identifying key moments in long videos is essential for downstream understanding and reasoning tasks. In this paper, we introduce a new problem, Taskoriented Temporal Grounding ToTG, which aims to localize time intervals containing the necessary information based on a task's natural description. Along with the definition, we also present ToTG Bench, a comprehensive benchmark for evaluating the performance on ToTG. ToTG is particularly challenging for traditional approaches due to their limited generalizability and difficulty in handling long videos. To address these challenges, we propose TimeScope, a novel framework built upon progressive reasoning. TimeScope first identifies a coarse-grained temporal scope in the long video that likely contains the key moments, and then refines this scope through finegrained moment partitioning. Additionally, we curate a highquality dataset, namely ToTG Pile, to enhance TimeScope's ability to perform progressive temporal grounding effectively. Extensive experiments demonstrate that TimeScope consistently outperforms both existing temporalgrounding methods and popular MLLMs across various settings, highlighting its effectiveness in addressing this new challenging problem.

Point tracking in videos is a fundamental task with applications in robotics, video editing, and more. While many vision tasks benefit from pre-trained feature backbones to improve generalizability, point tracking has primarily relied on simpler backbones trained from scratch on synthetic data, which may limit robustness in real-world scenarios. Additionally, point tracking requires temporal awareness to ensure coherence across frames, but using temporally-aware features is still underexplored. Most current methods often employ a two-stage process: an initial coarse prediction followed by a refinement stage to inject temporal information and correct errors from the coarse stage. These approach, however, is computationally expensive and potentially redundant if the feature backbone itself captures sufficient temporal information. In this work, we introduce Chrono, a feature backbone specifically designed for point tracking with built-in temporal awareness. Leveraging pre-trained representations from self-supervised learner DINOv2 and enhanced with a temporal adapter, Chrono effectively captures long-term temporal context, enabling precise prediction even without the refinement stage. Experimental results demonstrate that Chrono achieves state-of-the-art performance in a refiner-free setting on the TAP-Vid-DAVIS and TAP-Vid-Kinetics datasets, among common feature backbones used in point tracking as well as DINOv2, with exceptional efficiency. Project page: https://cvlab-kaist.github.io/Chrono/

We build upon time-series classification by leveraging the capabilities of Vision Language Models (VLMs). We find that VLMs produce competitive results after two or less epochs of fine-tuning. We develop a novel approach that incorporates graphical data representations as images in conjunction with numerical data. This approach is rooted in the hypothesis that graphical representations can provide additional contextual information that numerical data alone may not capture. Additionally, providing a graphical representation can circumvent issues such as limited context length faced by LLMs. To further advance this work, we implemented a scalable end-to-end pipeline for training on different scenarios, allowing us to isolate the most effective strategies for transferring learning capabilities from LLMs to Time Series Classification (TSC) tasks. Our approach works with univariate and multivariate time-series data. In addition, we conduct extensive and practical experiments to show how this approach works for time-series classification and generative labels.

Referring Video Object Segmentation (RefVOS) seeks to segment target objects in videos guided by natural language descriptions, demanding both temporal reasoning and fine-grained visual comprehension. Existing sampling strategies for LLM-based approaches typically rely on either handcrafted heuristics or external keyframe models. The former often overlooks essential temporal cues, while the latter increases system complexity. To address this, we propose a unified framework that jointly optimizes Temporal Sentence Grounding (TSG) and RefVOS, naturally incorporating key moment grounding capability. During training, we introduce a novel TSG paradigm that employs a dedicated [FIND] token for key moment identification through temporal token similarity matching, thereby avoiding the need for external timestamp encodings. For inference, we design a Moment-Centric Sampling (MCS) strategy that densely samples informative moments while sparsely sampling non-essential frames, preserving both motion details and global context. To further enhance tracking stability, we develop Bidirectional Anchor-updated Propagation (BAP), which leverages the most relevant moment as start point for high-quality mask initialization and dynamically updates at sampled points to mitigate accumulated errors. Code and model will be available at: https://github.com/Dmmm1997/MomentSeg

Current dialogue systems focus more on textual and speech context knowledge and are usually based on two speakers. Some recent work has investigated static image-based dialogue. However, several real-world human interactions also involve dynamic visual context (similar to videos) as well as dialogue exchanges among multiple speakers. To move closer towards such multimodal conversational skills and visually-situated applications, we introduce a new video-context, many-speaker dialogue dataset based on live-broadcast soccer game videos and chats from Twitch.tv. This challenging testbed allows us to develop visually-grounded dialogue models that should generate relevant temporal and spatial event language from the live video, while also being relevant to the chat history. For strong baselines, we also present several discriminative and generative models, e.g., based on tridirectional attention flow (TriDAF). We evaluate these models via retrieval ranking-recall, automatic phrase-matching metrics, as well as human evaluation studies. We also present dataset analyses, model ablations, and visualizations to understand the contribution of different modalities and model components.

Scene recognition is currently one of the top-challenging research fields in computer vision. This may be due to the ambiguity between classes: images of several scene classes may share similar objects, which causes confusion among them. The problem is aggravated when images of a particular scene class are notably different. Convolutional Neural Networks (CNNs) have significantly boosted performance in scene recognition, albeit it is still far below from other recognition tasks (e.g., object or image recognition). In this paper, we describe a novel approach for scene recognition based on an end-to-end multi-modal CNN that combines image and context information by means of an attention module. Context information, in the shape of semantic segmentation, is used to gate features extracted from the RGB image by leveraging on information encoded in the semantic representation: the set of scene objects and stuff, and their relative locations. This gating process reinforces the learning of indicative scene content and enhances scene disambiguation by refocusing the receptive fields of the CNN towards them. Experimental results on four publicly available datasets show that the proposed approach outperforms every other state-of-the-art method while significantly reducing the number of network parameters. All the code and data used along this paper is available at https://github.com/vpulab/Semantic-Aware-Scene-Recognition

Video generation has many unique challenges beyond those of image generation. The temporal dimension introduces extensive possible variations across frames, over which consistency and continuity may be violated. In this study, we move beyond evaluating simple actions and argue that generated videos should incorporate the emergence of new concepts and their relation transitions like in real-world videos as time progresses. To assess the Temporal Compositionality of video generation models, we propose TC-Bench, a benchmark of meticulously crafted text prompts, corresponding ground truth videos, and robust evaluation metrics. The prompts articulate the initial and final states of scenes, effectively reducing ambiguities for frame development and simplifying the assessment of transition completion. In addition, by collecting aligned real-world videos corresponding to the prompts, we expand TC-Bench's applicability from text-conditional models to image-conditional ones that can perform generative frame interpolation. We also develop new metrics to measure the completeness of component transitions in generated videos, which demonstrate significantly higher correlations with human judgments than existing metrics. Our comprehensive experimental results reveal that most video generators achieve less than 20% of the compositional changes, highlighting enormous space for future improvement. Our analysis indicates that current video generation models struggle to interpret descriptions of compositional changes and synthesize various components across different time steps.

We study the problem of generating temporal object intrinsics -- temporally evolving sequences of object geometry, reflectance, and texture, such as a blooming rose -- from pre-trained 2D foundation models. Unlike conventional 3D modeling and animation techniques that require extensive manual effort and expertise, we introduce a method that generates such assets with signals distilled from pre-trained 2D diffusion models. To ensure the temporal consistency of object intrinsics, we propose Neural Templates for temporal-state-guided distillation, derived automatically from image features from self-supervised learning. Our method can generate high-quality temporal object intrinsics for several natural phenomena and enable the sampling and controllable rendering of these dynamic objects from any viewpoint, under any environmental lighting conditions, at any time of their lifespan. Project website: https://chen-geng.com/rose4d

Modern image captioning system relies heavily on extracting knowledge from images to capture the concept of a static story. In this paper, we propose a textual visual context dataset for captioning, in which the publicly available dataset COCO Captions (Lin et al., 2014) has been extended with information about the scene (such as objects in the image). Since this information has a textual form, it can be used to leverage any NLP task, such as text similarity or semantic relation methods, into captioning systems, either as an end-to-end training strategy or a post-processing based approach.

Characters, locations, and actions recur throughout a movie, so many of its moments differ only in fine-grained details. Grounding such a moment in a full-length movie requires resolving that detail rather than matching the attributes it shares with other moments, yet existing temporal movie grounding benchmarks rarely test this. We introduce CoM^2-Bench, a benchmark of confusable query pairs, where the two queries in each pair share most attributes but differ in a single distinguishing detail within a full movie, posing a challenge for recent temporal grounding approaches. To locate where grounding breaks down, we analyze a retrieve-then-localize pipeline, in which a retriever first selects a short candidate chunk and a localizer then performs temporal grounding within it. Separating the two stages lets us attribute each failure to either retrieval or localization. We find that retrieval is the major bottleneck: state-of-the-art retrievers often select a moment's confusable counterpart, because their choices follow the content the two moments share rather than the detail that tells them apart. To address these failures, we propose a simple, training-free remedy, Agentic Window Search (AWS): an MLLM agent Decomposes the query into attribute-level sub-queries, then Inspects each candidate window against them and Expands the window when evidence is cut off at its edge. AWS works with existing retrievers and localizers without additional training, raising mIoU from 18.5 to 26.8.

Recent progress in using recurrent neural networks (RNNs) for image description has motivated the exploration of their application for video description. However, while images are static, working with videos requires modeling their dynamic temporal structure and then properly integrating that information into a natural language description. In this context, we propose an approach that successfully takes into account both the local and global temporal structure of videos to produce descriptions. First, our approach incorporates a spatial temporal 3-D convolutional neural network (3-D CNN) representation of the short temporal dynamics. The 3-D CNN representation is trained on video action recognition tasks, so as to produce a representation that is tuned to human motion and behavior. Second we propose a temporal attention mechanism that allows to go beyond local temporal modeling and learns to automatically select the most relevant temporal segments given the text-generating RNN. Our approach exceeds the current state-of-art for both BLEU and METEOR metrics on the Youtube2Text dataset. We also present results on a new, larger and more challenging dataset of paired video and natural language descriptions.

We propose Context-aware Video-text Alignment (CVA), a novel framework to address a significant challenge in video temporal grounding: achieving temporally sensitive video-text alignment that remains robust to irrelevant background context. Our framework is built on three key components. First, we propose Query-aware Context Diversification (QCD), a new data augmentation strategy that ensures only semantically unrelated content is mixed in. It builds a video-text similarity-based pool of replacement clips to simulate diverse contexts while preventing the ``false negative" caused by query-agnostic mixing. Second, we introduce the Context-invariant Boundary Discrimination (CBD) loss, a contrastive loss that enforces semantic consistency at challenging temporal boundaries, making their representations robust to contextual shifts and hard negatives. Third, we introduce the Context-enhanced Transformer Encoder (CTE), a hierarchical architecture that combines windowed self-attention and bidirectional cross-attention with learnable queries to capture multi-scale temporal context. Through the synergy of these data-centric and architectural enhancements, CVA achieves state-of-the-art performance on major VTG benchmarks, including QVHighlights and Charades-STA. Notably, our method achieves a significant improvement of approximately 5 points in Recall@1 (R1) scores over state-of-the-art methods, highlighting its effectiveness in mitigating false negatives.

Leveraging temporal context is crucial for success in partially observable robotic tasks. However, prior work in behavior cloning has demonstrated inconsistent performance gains when using multi-frame observations. In this paper, we introduce ContextVLA, a policy model that robustly improves robotic task performance by effectively leveraging multi-frame observations. Our approach is motivated by the key observation that Vision-Language-Action models (VLA), i.e., policy models built upon a Vision-Language Model (VLM), more effectively utilize multi-frame observations for action generation. This suggests that VLMs' inherent temporal understanding capability enables them to extract more meaningful context from multi-frame observations. However, the high dimensionality of video inputs introduces significant computational overhead, making VLA training and inference inefficient. To address this, ContextVLA compresses past observations into a single context token, allowing the policy to efficiently leverage temporal context for action generation. Our experiments show that ContextVLA consistently improves over single-frame VLAs and achieves the benefits of full multi-frame training but with reduced training and inference times.

Next-generation AI companions must go beyond general video understanding to resolve spatial and temporal references in dynamic, real-world environments. Existing Video Large Language Models (Video LLMs), while capable of coarse-level comprehension, struggle with fine-grained, spatiotemporal reasoning, especially when user queries rely on time-based event references for temporal anchoring, or gestural cues for spatial anchoring to clarify object references and positions. To bridge this critical gap, we introduce Strefer, a synthetic instruction data generation framework designed to equip Video LLMs with spatiotemporal referring and reasoning capabilities. Strefer produces diverse instruction-tuning data using a data engine that pseudo-annotates temporally dense, fine-grained video metadata, capturing rich spatial and temporal information in a structured manner, including subjects, objects, their locations as masklets, and their action descriptions and timelines. Our approach enhances the ability of Video LLMs to interpret spatial and temporal references, fostering more versatile, space-time-aware reasoning essential for real-world AI companions. Without using proprietary models, costly human annotation, or the need to annotate large volumes of new videos, experimental evaluations show that models trained with data produced by Strefer outperform baselines on tasks requiring spatial and temporal disambiguation. Additionally, these models exhibit enhanced space-time-aware reasoning, establishing a new foundation for perceptually grounded, instruction-tuned Video LLMs.

We address the problem of retrieving a specific moment from an untrimmed video by a query sentence. This is a challenging problem because a target moment may take place in relations to other temporal moments in the untrimmed video. Existing methods cannot tackle this challenge well since they consider temporal moments individually and neglect the temporal dependencies. In this paper, we model the temporal relations between video moments by a two-dimensional map, where one dimension indicates the starting time of a moment and the other indicates the end time. This 2D temporal map can cover diverse video moments with different lengths, while representing their adjacent relations. Based on the 2D map, we propose a Temporal Adjacent Network (2D-TAN), a single-shot framework for moment localization. It is capable of encoding the adjacent temporal relation, while learning discriminative features for matching video moments with referring expressions. We evaluate the proposed 2D-TAN on three challenging benchmarks, i.e., Charades-STA, ActivityNet Captions, and TACoS, where our 2D-TAN outperforms the state-of-the-art.

Diverse actions give rise to rich audio-visual signals in long videos. Recent works showcase that the two modalities of audio and video exhibit different temporal extents of events and distinct labels. We address the interplay between the two modalities in long videos by explicitly modelling the temporal extents of audio and visual events. We propose the Time Interval Machine (TIM) where a modality-specific time interval poses as a query to a transformer encoder that ingests a long video input. The encoder then attends to the specified interval, as well as the surrounding context in both modalities, in order to recognise the ongoing action. We test TIM on three long audio-visual video datasets: EPIC-KITCHENS, Perception Test, and AVE, reporting state-of-the-art (SOTA) for recognition. On EPIC-KITCHENS, we beat previous SOTA that utilises LLMs and significantly larger pre-training by 2.9% top-1 action recognition accuracy. Additionally, we show that TIM can be adapted for action detection, using dense multi-scale interval queries, outperforming SOTA on EPIC-KITCHENS-100 for most metrics, and showing strong performance on the Perception Test. Our ablations show the critical role of integrating the two modalities and modelling their time intervals in achieving this performance. Code and models at: https://github.com/JacobChalk/TIM