论文检索

Haodong Yan, Hang Yu, Zhide Zhong, Weilin Yuan, Xin Gong, Zehang Luo, Chengxi Heyu, Junfeng Li, Wenxuan Song, Shunbo Zhou 等

Generating realistic hand-object interactions (HOI) videos is a significant challenge due to the difficulty of modeling physical constraints (e.g., contact and occlusion between hands and manipulated objects). Current methods utilize HOI representation as an auxiliary generative objective to guide video synthesis. However, there is a dilemma between 2D and 3D representations that cannot simultaneously guarantee scalability and interaction fidelity. To address this limitation, we propose a structure and contact-aware representation that captures hand-object contact, hand-object occlusion, and holistic structure context without 3D annotations. This interaction-oriented and scalable supervision signal enables the model to learn fine-grained interaction physics and generalize to open-world scenarios. To fully exploit the proposed representation, we introduce a joint-generation paradigm with a share-and-specialization strategy that generates interaction-oriented representations and videos. Extensive experiments demonstrate that our method outperforms state-of-the-art methods on two real-world datasets in generating physics-realistic and temporally coherent HOI videos. Furthermore, our approach exhibits strong generalization to challenging open-world scenarios, highlighting the benefit of our scalable design.

Gu Zhang, Qicheng Xu, Haozhe Zhang, Jianhan Ma, Long He, Yiming Bao, Zeyu Ping, Zhecheng Yuan, Chenhao Lu, Chengbo Yuan 等

Dexterous manipulation remains challenging due to the cost of collecting real-robot teleoperation data, the heterogeneity of hand embodiments, and the high dimensionality of control. We present UniDex, a robot foundation suite that couples a large-scale robot-centric dataset with a unified vision-language-action (VLA) policy and a practical human-data capture setup for universal dexterous hand control. First, we construct UniDex-Dataset, a robot-centric dataset over 50K trajectories across eight dexterous hands (6-24 DoFs), derived from egocentric human video datasets. To transform human data into robot-executable trajectories, we employ a human-in-the-loop retargeting procedure to align fingertip trajectories while preserving plausible hand-object contacts, and we operate on explicit 3D pointclouds with human hands masked to narrow kinematic and visual gaps. Second, we introduce the Function-Actuator-Aligned Space (FAAS), a unified action space that maps functionally similar actuators to shared coordinates, enabling cross-hand transfer. Leveraging FAAS as the action parameterization, we train UniDex-VLA, a 3D VLA policy pretrained on UniDex-Dataset and finetuned with task demonstrations. In addition, we build UniDex-Cap, a simple portable capture setup that records synchronized RGB-D streams and human hand poses and converts them into robot-executable trajectories to enable human-robot data co-training that reduces reliance on costly robot demonstrations. On challenging tool-use tasks across two different hands, UniDex-VLA achieves 81% average task progress and outperforms prior VLA baselines by a large margin, while exhibiting strong spatial, object, and zero-shot cross-hand generalization. Together, UniDex-Dataset, UniDex-VLA, and UniDex-Cap provide a scalable foundation suite for universal dexterous manipulation.

Aoyu Liu, Zhen Liu, Ziyi Wang, Dian Chen, Bing Zeng, Shuaicheng Liu

Single-image HDR reconstruction aims to recover high dynamic range radiance from a single low dynamic range (LDR) input, but remains highly ill-posed due to detail saturation in over-exposed regions and noise amplification in under-exposed areas. While recent diffusion-based approaches offer powerful generative priors, they often overlook the exposure-dependent nature of the degradation and incur substantial computational costs from iterative sampling. To address these challenges, we propose ExpoCM, a novel one-step generative HDR reconstruction framework that reformulates HDR reconstruction as a Probability Flow ODE (PF-ODE) and constructs exposure-aware consistency trajectories via exposure-dependent perturbations. Specifically, a soft exposure mask is first constructed to separate the LDR image into over-, under-, and well-exposed regions. Based on this partition, region-conditioned consistency trajectories are designed to hallucinate saturated details, suppress noise in dark regions, and preserve reliable structures within a single, distillation-free inference step. To further enhance perceptual quality, we introduce an Exposure-guided Luminance-Chromaticity Loss in the CIE \text L ^*\text a ^*\text b ^* space, which assigns exposure-aware weights to luminance and chromaticity components, effectively mitigating brightness bias and color drift. Extensive experiments on the HDR-REAL, HDR-EYE, and AIM2025 benchmarks demonstrate that ExpoCM achieves state-of-the-art fidelity and perceptual accuracy, while enabling over 400xand 20xfaster inference compared to DDPM (1000 steps) and DDIM (50 steps), respectively. Code is available at https://github.com/AoyuLiu01/ExpoCM.

Chenfan Qu, Yiwu Zhong, Xuekang Zhu, Junchi Li, Changjiang Jiang, Jian liu, Lianwen Jin

The rapid advancement of generative AI enables the creation of highly realistic text images, posing significant security risks from fraud and disinformation. However, research into robust detection is critically hampered by existing datasets that lack scale, diversity, and updated counterfeit techniques, as well as by models that fail to generalize. To address these deficiencies, we introduce DanceText, a large-scale, comprehensive dataset for AI-counterfeited text image detection. Constructed using our novel Creative Proposer pipeline, which automates the generation of diverse and realistic counterfeits, DanceText surpasses previous works by over 100-fold in multiple dimensions. It is the first to include counterfeits from multimodal large models, commercial software, and mobile apps, covering all major paradigms, including full-image generation, regional removal, and editing. Building on this dataset, we propose DS-Net, a novel and effective detection model. It features two key components: a Forensic Decoupling Encoder to extract generator-agnostic artifact features, and a Synergy Denoising Decoder that synergizes image-level classification with instance-level localization. Extensive experiments demonstrate that DS-Net achieves state-of-the-art performance, advancing the field toward robust and unified detection of AI-counterfeited text images in real-world scenarios. Both our code and dataset will be released publicly.

Dingkun Yan, Xinrui Wang, Ru Wang, Zhuoru Li, Jinze Yu, Yusuke Iwasawa, Yutaka Matsuo, Jiaxian Guo

Sketch colorization models have been widely studied to automate and assist in the creation of animation frames and digital illustrations. However, current methods are still not satisfactory for industrial standard applications in high-resolution synthesis and precise controllability of details. To further enhance the synthesis quality and controllability, we propose an image-referenced sketch colorization method based on the powerful SDXL backbone and leverage sketches as spatial guidance and RGB images as color references. A split cross-attention mechanism is coupled with spatial masks to separately colorize the foreground and background regions to avoid spatial entanglement. A tagger network trained on a massive anime-style image dataset is employed to extract attribution-level information from reference images and integrated into the pipeline to provide precise control signals for synthesis. However, the increased resolution and number of attention layers in the SDXL backbone and precise reference information from the tagger network cause severe entanglement during colorization. We consequently combine a foreground encoder and a background encoder for disentanglement and better synthesis quality. Furthermore, a high-quality annotated and paired sketch colorization dataset is collected for fine-tuning. The proposed method is the first to achieve high resolution high quality sketch colorization with precise control, and obvious outperforms existing methods in quantitative and qualitative validations, as well as user studies in both quality and controllability. Ablation study reveals the influence of each component. Code and dataset will be made publicly available upon paper acceptance.

Tianling Xu, Shengzhe Gan, Leslie Gu, Yuelei Li, Fangneng Zhan, Hanspeter Pfister

Active 3D reconstruction enables an agent to autonomously select viewpoints to build accurate and complete scene geometry efficiently, rather than passively reconstructing scenes from pre-collected images. Existing active reconstruction methods often rely on geometric heuristics, which may result in redundant observations without improving reconstruction quality. To address this, we propose AREA3D, an active reconstruction agent for 3D reconstruction by leveraging feed-forward 3D models and vision-language guidance. The framework decouples view uncertainty modeling from feed-forward reconstruction, enabling precise uncertainty estimation without online optimization. Moreover, the integrated Vision-Language Model provides high-level semantic guidance that guides exploration beyond purely geometric cues. Extensive experiments on both scene-level and object-level benchmarks (Replica and OmniObject3D) demonstrate that AREA3D achieves state-of-the-art reconstruction accuracy, especially in sparse views.

Derong Jin, Xiyi Chen, Ming C. Lin, Ruohan Gao

Tremendous progress in visual scene generation now turns a single image into an explorable 3D world, yet immersion remains incomplete without sound. We introduce Image2AVScene, the task of generating a 3D audio-visual scene from a single image, and present SonoWorld, the first framework to tackle this challenge. From one image, our pipeline outpaints a 360deg panorama, lifts it into a navigable 3D scene, places language-guided sound anchors, and renders ambisonics for point, areal, and ambient sources, yielding spatial audio aligned with scene geometry and semantics. Quantitative evaluations on a newly curated real-world dataset and a controlled user study confirm the effectiveness of our approach. Beyond free-viewpoint audio-visual rendering, we also demonstrate applications to one-shot acoustic learning and audio-visual spatial source separation. Project website: https://humathe.github.io/sonoworld/

Zehua Zang, Xi Wang, Fuchun Sun, Xiao Xu, Lixiang Liu, Jiahuan Zhou, Jiangmeng Li

Vision-Language-Action models (VLAs) achieve remarkable performance in sequential decision-making but remain fragile to subtle environmental shifts, such as small changes in object pose. We attribute this brittleness to trajectory overfitting, where VLAs over-attend to the spurious correlation between actions and entities, then reproduce memorized action patterns. We propose Perturbation learning with Delayed Feedback (PDF), a verifier-free test-time adaptation framework that improves decision performance without fine-tuning the base model. PDF mitigates the spurious correlation through uncertainty-based data augmentation and action voting, while an adaptive scheduler allocates augmentation budgets to balance performance and efficiency. To further improve stability, PDF learns a lightweight perturbation module that retrospectively adjusts action logits guided by delayed feedback, correcting overconfidence issue. Experiments on LIBERO (+7.4% success rate) and Atari (+10.3 human normalized score) demonstrate consistent gains of PDF in task success over vanilla VLA and VLA with test-time adaptation, establishing a practical path toward reliable test-time adaptation in multimodal decision-making agents.

Jiayao Tan, Fan Lyu, Tianle Liu, Fuyuan Hu, Wei Feng

Multimodal Continual Learning (MMCL) aims to enable models to continuously accumulate knowledge across multiple tasks and modalities without forgetting prior information. MMCL presents more challenges than single-modal continual learning, as it requires effective cooperation and complementarity between modalities. Existing methods often treat modality alignment as a static process, assuming once alignment is established, it remains fixed. However, we argue that modality alignment is inherently dynamic, evolving with task learning and feature propagation across layers. To address this, we introduce Dynamic Alignment Graph Regularization (DAGR), a novel approach that explicitly models the evolving alignment across layers. By incorporating multi-level graph regularization, our method stabilizes the alignment process and mitigates catastrophic forgetting. Extensive experiments on benchmarks, such as MTIL, show that DAGR outperforms static alignment-based methods and other continual learning techniques, achieving superior stability.

Bin Li, Ruichi Zhang, Han Liang, Jingyan Zhang, Juze Zhang, Xin Chen, Lan Xu, Jingyi Yu, Jingya Wang

Humanoid agents are expected to emulate the complex coordination inherent in human social behaviors. However, existing methods are largely confined to single-agent scenarios, overlooking the physically plausible interplay essential for multi-agent interactions. To bridge this gap, we propose InterAgent, the first end-to-end framework for text-driven physics-based multi-agent humanoid control. At its core, we introduce an autoregressive diffusion transformer equipped with multi-stream blocks, which decouples proprioception, exteroception, and action to mitigate cross-modal interference while enabling synergistic coordination. We further propose a novel interaction graph exteroception representation that explicitly captures fine-grained joint-to-joint spatial dependencies to facilitate network learning. Additionally, within it we devise a sparse edge-based attention mechanism that dynamically prunes redundant connections and emphasizes critical inter-agent spatial relations, thereby enhancing the robustness of interaction modeling. Extensive experiments demonstrate that InterAgent consistently outperforms multiple strong baselines, achieving state-of-the-art performance. It enables producing coherent, physically plausible, and semantically faithful multi-agent behaviors from only text prompts. Project page: \tt \small \href https://binlee26.github.io/InterAgent-Page https://binlee26.github.io/InterAgent-Page .

Yongsung Kim, Wooseok Song, Jaihyun Lew, Hun Hwangbo, Jaehoon Lee, Sungroh Yoon

Visual Geometry Grounded Transformer (VGGT) has advanced 3D vision, yet its global attention layers suffer from quadratic computational costs that hinder scalability. Several sparsification-based acceleration techniques have been proposed to alleviate this issue, but they often suffer from substantial accuracy degradation. We hypothesize that the accuracy degradation stems from the heterogeneity in head-wise sparsification sensitivity, as the existing methods apply a uniform sparsity pattern across all heads. Motivated by this hypothesis, we present a two-stage sparsification pipeline that effectively quantifies and exploits headwise sparsification sensitivity. In the first stage, we measure head-wise sparsification sensitivity using a novel metric, the Head Sensitivity Score (HeSS), which approximates the Hessian with respect to two distinct error terms on a small calibration set. In the inference stage, we perform HeSS-Guided Sparsification, leveraging the pre-computed HeSS to reallocate the total attention budget-assigning denser attention to sensitive heads and sparser attention to more robust ones. We demonstrate that HeSS effectively captures head-wise sparsification sensitivity and empirically confirm that attention heads in the global attention layers exhibit heterogeneous sensitivity characteristics. Extensive experiments further show that our method effectively mitigates performance degradation under high sparsity, demonstrating strong robustness across varying sparsification levels. Code is available at https://github.com/libary753/HeSS.

Taejin Jeong, Joohyeok Kim, Jinyeong Kim, Chanyoung Kim, Seong Jae Hwang

Spatial Transcriptomics (ST) provides spatially-resolved gene expression, offering crucial insights into tissue architecture and complex diseases. However, its prohibitive cost limits widespread adoption, leading to significant attention on inferring spatial gene expression from readily available whole slide images. While graph neural networks have been proposed to model interactions between tissue regions, their reliance on pre-defined sparse graphs prevents them from considering potentially interacting spot pairs, resulting in a structural limitation in capturing complex biological relationships. To address this, we propose FEAST (Fully connected Expressive Attention for Spatial Transcriptomics), an attention-based framework that models the tissue as a fully connected graph, enabling the consideration of all pairwise interactions. To better reflect biological interactions, we introduce negative-aware attention, which models both excitatory and inhibitory interactions, capturing essential negative relationships that standard attention often overlooks. Furthermore, to mitigate the information loss from truncated or ignored context in standard spot image extraction, we introduce an off-grid sampling strategy that gathers additional images from intermediate regions, allowing the model to capture a richer morphological context. Experiments on public ST datasets show that FEAST surpasses state-of-the-art methods in gene expression prediction while providing biologically plausible attention maps that clarify positive and negative interactions. Our code is available at https://github.com/starforTJ/ FEAST.

Yuchen Wu, Jiahe Li, Xiaohan Yu, Lina Yu, Jin Zheng, Xiao Bai

Monocular visual SLAM enables 3D reconstruction from internet video and autonomous navigation on resource-constrained platforms, yet suffers from scale drift, i.e., the gradual divergence of estimated scale over long sequences. Existing frame-to-frame methods achieve real-time performance through local optimization but accumulate scale drift due to the lack of global constraints among independent windows. To address this, we propose SCE-SLAM, an end-to-end SLAM system that maintains scale consistency through scene coordinate embeddings, which are learned patch-level representations encoding 3D geometric relationships under a canonical scale reference. The framework consists of two key modules: geometry-guided aggregation that leverages 3D spatial proximity to propagate scale information from historical observations through geometry-modulated attention, and scene coordinate bundle adjustment that anchors current estimates to the reference scale through explicit 3D coordinate constraints decoded from the scene coordinate embeddings. Experiments on KITTI, Waymo, and vKITTI demonstrate substantial improvements: our method reduces absolute trajectory error by 8.36m on KITTI compared to the best prior approach, while maintaining 36 FPS and achieving scale consistency across large-scale scenes.

Chenghao Li, Jun Liu, Songbo Zhang, Huadong Jian, Hao Ni, Lik-Hang Lee, Sung-Ho Bae, Guoqing Wang, Yang Yang, Chaoning Zhang

Multimodal LLM agents operating in complex game environments must continually reuse past experience to solve new tasks efficiently. In this work, we propose Echo, a transfer-oriented memory framework that enables agents to derive actionable knowledge from prior interactions rather than treating memory as a passive repository of static records. To make transfer explicit, Echo decomposes reusable knowledge into five dimensions: structure, attribute, process, function, and interaction. This formulation allows the agent to identify recurring patterns shared across different tasks and infer what prior experience remains applicable in new situations. Building on this formulation, Echo leverages In-Context Analogy Learning (ICAL) to retrieve relevant experiences and adapt them to unseen tasks through contextual examples. Experiments in Minecraft show that, under a from-scratch learning setting, Echo achieves a 1.3x-1.7x speed-up on object-unlocking tasks. Moreover, Echo exhibits a burst-like chain-unlocking phenomenon, rapidly unlocking multiple similar items within a short time interval after acquiring transferable experience. These results suggest that experience transfer is a promising direction for improving the efficiency and adaptability of multimodal LLM agents in complex interactive environments.

Yakun Chang, Zhaojun Huang, Siqi Yang, Yeliduosi Xiaokaiti, Shikui Wei, Yao Zhao, Tiejun Huang, Boxin Shi

Capturing scenes with both high dynamic range (HDR) and high-speed motion remains challenging for conventional cameras. Existing alternating-exposure approaches exacerbate temporal resolution loss, making them unsuitable for high-speed scenes. Consequently, current solutions typically compromise either spatial resolution through fixed spatial-varying attenuation levels or employ multi-sensor configurations to maintain temporal resolution. In this paper, we leverage an ultra-high speed spike camera to enable spatial and temporal attenuation of incident light, thereby reconstructing high frame rate (HFR) and HDR video with a single sensor. We achieve this by placing a rapidly rotating spoke-pattern neutral density (SpokeND) filter in front of the spike camera, enabling each pixel to periodically capture multi-attenuated spikes while maintaining full spatial resolution. Building on these multi-attenuated spikes, we propose ReST-Net, which comprises the ReGain and ReFine modules. The ReGain module reconstructs spatially consistent frames by learning to recover relative gain from the multi-attenuated spikes, and the ReFine module removes temporal fluctuations to produce temporally consistent HDR videos. Extensive experiments on synthetic and real-world data demonstrate that our method can reconstruct HDR video at up to 2000 FPS.

Xuan Lu, Kangle Li, Haohang Huang, Rui Meng, Wenjun Zeng, Xiaoyu Shen

Recent advances in multimodal large language models (MLLMs) have substantially expanded the capabilities of multimodal retrieval, enabling systems to align and retrieve information across visual and textual modalities. Yet, existing benchmarks largely focus on coarse-grained or single-condition alignment, overlooking real-world scenarios where user queries specify multiple interdependent constraints across modalities. To bridge this gap, we introduce MCMR (Multi-Conditional Multimodal Retrieval): a large-scale benchmark designed to evaluate fine-grained, multi-condition cross-modal retrieval under natural-language queries. MCMR spans five product domains: upper and bottom clothing, jewelry, shoes, and furniture. It also preserves rich long-form metadata essential for compositional matching. Each query integrates complementary visual and textual attributes, requiring models to jointly satisfy all specified conditions for relevance. We benchmark a diverse suite of MLLM-based multimodal retrievers and vision-language rerankers to assess their condition-aware reasoning abilities. Experimental results reveal: (i) distinct modality asymmetries across models; (ii) visual cues dominate early-rank precision, while textual metadata stabilizes long-tail ordering; and (iii) MLLM-based pointwise rerankers markedly improve fine-grained matching by explicitly verifying query-candidate consistency. Overall, MCMR establishes a challenging and diagnostic benchmark for advancing multimodal retrieval toward compositional, constraint-aware, and interpretable understanding. Our code and dataset is available at https://github.com/EIT-NLP/MCMR

Alexis Jensen, Pei Xu, Ioannis Karamouzas, Charles Pontonnier, Julien Pettré

We present a physics-based method for simulating full-body agents that recover balance by stepping or applying contact forces after being perturbed in dense crowds. While traditional 2D crowd simulations focus on navigation and social interactions in moderately dense settings, interactions in highly dense environments are predominantly physical, leading to push propagation, falls, and potential hazards. Existing models cannot capture how forces are transmitted through the body at the limb level. To address this, we use physics-based anthropomorphic simulations combined with a two-stage deep reinforcement learning framework. In the first stage, a policy is pre-trained using reference motion data and general balance rewards, enabling agents to handle a wide range of perturbations. In the second stage, an adaptive phase refines the policy to allow socially aware interactions, using hand contacts for stabilization guided by an online heuristic targeting neighbors' shoulders based on mechanical efficiency and collision risk. Ablation studies validate the training framework and reward components, and simulations reproduce trends observed in empirical studies of push propagation. Our method scales to large populations, offering new opportunities to study safety and collective behavior in dense crowds.

Stefanos Koutoupis, Michaela Areti Zervou, Konstantinos Kontras, Maarten De Vos, Panagiotis Tsakalides, Grigorios Tsagkatakis

Learning joint representations across multiple modalities remains a central challenge in multimodal machine learning. Prevailing approaches predominantly operate in pairwise settings, aligning two modalities at a time. While some recent methods aim to capture higher-order interactions among multiple modalities, they often overlook or insufficiently preserve pairwise relationships, limiting their effectiveness on single-modality tasks. In this work, we introduce Contrastive Fusion (ConFu), a framework that jointly embeds both individual modalities and their fused combinations into a unified representation space, where modalities and their fused counterparts are aligned. ConFu extends traditional pairwise contrastive objectives with an additional fused-modality contrastive term, encouraging the joint embedding of modality pairs with a third modality. This formulation enables ConFu to capture higher-order dependencies, such as XOR-like relationships, that cannot be recovered through pairwise alignment alone, while still maintaining strong pairwise correspondence. We evaluate ConFu on synthetic and real-world multimodal benchmarks, assessing its ability to exploit cross-modal complementarity, capture higher-order dependencies, and scale with increasing multimodal complexity. Across these settings, ConFu demonstrates competitive performance on retrieval and classification tasks, while supporting unified one-to-one and two-to-one retrieval within a single contrastive framework. We release our code and dataset at https://github.com/estafons/confu.

Xiaokun Sun, Zeyu Cai, Hao Tang, Ying Tai, Jian Yang, Zhenyu Zhang

3D morphing remains challenging due to the difficulty of generating semantically consistent and temporally smooth deformations, especially across categories. We present MorphAny3D, a training-free framework that leverages Structured Latent (SLAT) representations for high-quality 3D morphing. Our key insight is that intelligently blending source and target SLAT features within the attention mechanisms of 3D generators naturally produces plausible morphing sequences. To this end, we introduce Morphing Cross-Attention (MCA), which fuses source and target information for structural coherence, and Temporal-Fused Self-Attention (TFSA), which enhances temporal consistency by incorporating features from preceding frames. An orientation correction strategy further mitigates the pose ambiguity within the morphing steps. Extensive experiments show that our method generates state-of-the-art morphing sequences, even for challenging cross-category cases. MorphAny3D further supports advanced applications such as decoupled morphing and 3D style transfer, and can be generalized to other SLAT-based generative models. Project page: https://xiaokunsun.github.io/MorphAny3D.github.io/.

Shengchao Zhou, Yuxin Chen, Yuying Ge, Wei Huang, Jiehong Lin, Ying Shan, Xiaojuan Qi

Vision-language models (VLM) excel at general understanding yet remain weak at dynamic spatial reasoning (DSR), i.e., reasoning about the evolvement of object geometry and relationship in 3D space over time, largely due to the scarcity of scalable 4D-aware training resources. To bridge this gap across aspects of dataset, benchmark and model, we introduce DSR Suite. First, we propose an automated pipeline that generates multiple-choice question-answer pairs from in-the-wild videos for DSR. By leveraging modern vision foundation models, the pipeline extracts rich geometric and motion information, including camera poses, local point clouds, object masks, orientations, and 3D trajectories. These geometric cues enable the construction of DSR-Train for learning and further human-refined DSR-Bench for evaluation. Compared with previous works, our data emphasize (i) in-the-wild video sources, (ii) object- and scene-level 3D requirements, (iii) viewpoint transformations, (iv) multi-object interactions, and (v) fine-grained, procedural answers. Beyond data, we propose a lightweight Geometry Selection Module (GSM) to seamlessly integrate geometric priors into VLMs, which condenses question semantics and extracts question-relevant knowledge from pretrained 4D reconstruction priors into a compact set of geometry tokens. This targeted extraction avoids overwhelming the model with irrelevant knowledge. Experiments show that integrating DSR-Train and GSM into Qwen2.5-VL-7B significantly enhances its dynamic spatial reasoning capability, while maintaining accuracy on general video understanding benchmarks.