arXiv subjects

Limin Wang

Publications and source records attributed to Limin Wang.

At least 91 records · Page 5Linked to original sources

arXiv2025

Weakly Supervised Temporal Sentence Grounding via Positive Sample Mining

The task of weakly supervised temporal sentence grounding (WSTSG) aims to detect temporal intervals corresponding to a language description from untrimmed videos with only video-level video-language correspondence. For an anchor sample, most existing approaches generate negative samples either from other videos or within the same video for contrastive learning. However, some training samples are highly similar to the anchor sample, directly regarding them as negative samples leads to difficulties for optimization and ignores the correlations between these similar samples and the anchor sample. To address this, we propose Positive Sample Mining (PSM), a novel framework that mines positive samples from the training set to provide more discriminative supervision. Specifically, for a given anchor sample, we partition the remaining training set into semantically similar and dissimilar subsets based on the similarity of their text queries. To effectively leverage these correlations, we introduce a PSM-guided contrastive loss to ensure that the anchor proposal is closer to similar samples and further from dissimilar ones. Additionally, we design a PSM-guided rank loss to ensure that similar samples are closer to the anchor proposal than to the negative intra-video proposal, aiming to distinguish the anchor proposal and the negative intra-video proposal. Experiments on the WSTSG and grounded VideoQA tasks demonstrate the effectiveness and superiority of our method.

cs.CV↗

arXiv2025

Joint Modeling of Feature, Correspondence, and a Compressed Memory for Video Object Segmentation

Current prevailing Video Object Segmentation methods follow the pipeline of extraction-then-matching, which first extracts features on current and reference frames independently, and then performs dense matching between them. This decoupled pipeline limits information propagation between frames to high-level features, hindering fine-grained details for matching. Furthermore, the pixel-wise matching lacks holistic target understanding, making it prone to disturbance by similar distractors. To address these issues, we propose a unified VOS framework, coined as JointFormer, for jointly modeling feature extraction, correspondence matching, and a compressed memory. The core Joint Modeling Block leverages attention to simultaneously extract and propagate the target information from the reference frame to the current frame and a compressed memory token. This joint scheme enables extensive multi-layer propagation beyond high-level feature space and facilitates robust instance-distinctive feature learning. To incorporate the long-term and holistic target information, we introduce a compressed memory token with a customized online updating mechanism, which aggregates target features and facilitates temporal information propagation in a frame-wise manner, enhancing global modeling consistency. Our JointFormer achieves a new state-of-the-art performance on the DAVIS 2017 val/test-dev (89.7\% and 87.6\%) benchmarks and the YouTube-VOS 2018/2019 val (87.0\% and 87.0\%) benchmarks, outperforming the existing works. To demonstrate the generalizability of our model, it is further evaluated on four new benchmarks with various difficulties, including MOSE for complex scenes, VISOR for egocentric videos, VOST for complex transformations, and LVOS for long-term videos.

cs.CV↗

arXiv2025

Eagle 2.5: Boosting Long-Context Post-Training for Frontier Vision-Language Models

We introduce Eagle 2.5, a family of frontier vision-language models (VLMs) for long-context multimodal learning. Our work addresses the challenges in long video comprehension and high-resolution image understanding, introducing a generalist framework for both tasks. The proposed training framework incorporates Automatic Degrade Sampling and Image Area Preservation, two techniques that preserve contextual integrity and visual details. The framework also includes numerous efficiency optimizations in the pipeline for long-context data training. Finally, we propose Eagle-Video-110K, a novel dataset that integrates both story-level and clip-level annotations, facilitating long-video understanding. Eagle 2.5 demonstrates substantial improvements on long-context multimodal benchmarks, providing a robust solution to the limitations of existing VLMs. Notably, our best model Eagle 2.5-8B achieves 72.4% on Video-MME with 512 input frames, matching the results of top-tier commercial model such as GPT-4o and large-scale open-source models like Qwen2.5-VL-72B and InternVL2.5-78B.

cs.CV↗

arXiv2025

InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models

We introduce InternVL3, a significant advancement in the InternVL series featuring a native multimodal pre-training paradigm. Rather than adapting a text-only large language model (LLM) into a multimodal large language model (MLLM) that supports visual inputs, InternVL3 jointly acquires multimodal and linguistic capabilities from both diverse multimodal data and pure-text corpora during a single pre-training stage. This unified training paradigm effectively addresses the complexities and alignment challenges commonly encountered in conventional post-hoc training pipelines for MLLMs. To further improve performance and scalability, InternVL3 incorporates variable visual position encoding (V2PE) to support extended multimodal contexts, employs advanced post-training techniques such as supervised fine-tuning (SFT) and mixed preference optimization (MPO), and adopts test-time scaling strategies alongside an optimized training infrastructure. Extensive empirical evaluations demonstrate that InternVL3 delivers superior performance across a wide range of multi-modal tasks. In particular, InternVL3-78B achieves a score of 72.2 on the MMMU benchmark, setting a new state-of-the-art among open-source MLLMs. Its capabilities remain highly competitive with leading proprietary models, including ChatGPT-4o, Claude 3.5 Sonnet, and Gemini 2.5 Pro, while also maintaining strong pure-language proficiency. In pursuit of open-science principles, we will publicly release both the training data and model weights to foster further research and development in next-generation MLLMs.

cs.CV↗

arXiv2025

Online Video Understanding: OVBench and VideoChat-Online

Multimodal Large Language Models (MLLMs) have significantly progressed in offline video understanding. However, applying these models to real-world scenarios, such as autonomous driving and human-computer interaction, presents unique challenges due to the need for real-time processing of continuous online video streams. To this end, this paper presents systematic efforts from three perspectives: evaluation benchmark, model architecture, and training strategy. First, we introduce OVBench, a comprehensive question-answering benchmark designed to evaluate models' ability to perceive, memorize, and reason within online video contexts. It features 6 core task types across three temporal contexts-past, current, and future-forming 16 subtasks from diverse datasets. Second, we propose a new Pyramid Memory Bank (PMB) that effectively retains key spatiotemporal information in video streams. Third, we proposed an offline-to-online learning paradigm, designing an interleaved dialogue format for online video data and constructing an instruction-tuning dataset tailored for online video training. This framework led to the development of VideoChat-Online, a robust and efficient model for online video understanding. Despite the lower computational cost and higher efficiency, VideoChat-Online outperforms existing state-of-the-art offline and online models across popular offline video benchmarks and OVBench, demonstrating the effectiveness of our model architecture and training strategy. % Our approach surpasses existing state-of-the-art offline models Qwen2-VL 7B and online models Flash-VStream, by 4.19% and 23.7% on OVBench, respectively.

cs.CV↗

arXiv2025

DMM: Building a Versatile Image Generation Model via Distillation-Based Model Merging

The success of text-to-image (T2I) generation models has spurred a proliferation of numerous model checkpoints fine-tuned from the same base model on various specialized datasets. This overwhelming specialized model production introduces new challenges for high parameter redundancy and huge storage cost, thereby necessitating the development of effective methods to consolidate and unify the capabilities of diverse powerful models into a single one. A common practice in model merging adopts static linear interpolation in the parameter space to achieve the goal of style mixing. However, it neglects the features of T2I generation task that numerous distinct models cover sundry styles which may lead to incompatibility and confusion in the merged model. To address this issue, we introduce a style-promptable image generation pipeline which can accurately generate arbitrary-style images under the control of style vectors. Based on this design, we propose the score distillation based model merging paradigm (DMM), compressing multiple models into a single versatile T2I model. Moreover, we rethink and reformulate the model merging task in the context of T2I generation, by presenting new merging goals and evaluation protocols. Our experiments demonstrate that DMM can compactly reorganize the knowledge from multiple teacher models and achieve controllable arbitrary-style generation.

cs.CV↗

arXiv2025

Contextual AD Narration with Interleaved Multimodal Sequence

The Audio Description (AD) task aims to generate descriptions of visual elements for visually impaired individuals to help them access long-form video content, like movies. With video feature, text, character bank and context information as inputs, the generated ADs are able to correspond to the characters by name and provide reasonable, contextual descriptions to help audience understand the storyline of movie. To achieve this goal, we propose to leverage pre-trained foundation models through a simple and unified framework to generate ADs with interleaved multimodal sequence as input, termed as Uni-AD. To enhance the alignment of features across various modalities with finer granularity, we introduce a simple and lightweight module that maps video features into the textual feature space. Moreover, we also propose a character-refinement module to provide more precise information by identifying the main characters who play more significant roles in the video context. With these unique designs, we further incorporate contextual information and a contrastive loss into our architecture to generate smoother and more contextually appropriate ADs. Experiments on multiple AD datasets show that Uni-AD performs well on AD generation, which demonstrates the effectiveness of our approach. Our code is available at: https://github.com/ant-research/UniAD.

cs.CV↗

arXiv2025

Shrinkage Initialization for Smooth Learning of Neural Networks

The successes of intelligent systems have quite relied on the artificial learning of information, which lead to the broad applications of neural learning solutions. As a common sense, the training of neural networks can be largely improved by specifically defined initialization, neuron layers as well as the activation functions. Though there are sequential layer based initialization available, the generalized solution to initial stages is still desired. In this work, an improved approach to initialization of neural learning is presented, which adopts the shrinkage approach to initialize the transformation of each layer of networks. It can be universally adapted for the structures of any networks with random layers, while stable performance can be attained. Furthermore, the smooth learning of networks is adopted in this work, due to the diverse influence on neural learning. Experimental results on several artificial data sets demonstrate that, the proposed method is able to present robust results with the shrinkage initialization, and competent for smooth learning of neural networks.

cs.LG↗

arXiv2025

DDT: Decoupled Diffusion Transformer

Diffusion transformers have demonstrated remarkable generation quality, albeit requiring longer training iterations and numerous inference steps. In each denoising step, diffusion transformers encode the noisy inputs to extract the lower-frequency semantic component and then decode the higher frequency with identical modules. This scheme creates an inherent optimization dilemma: encoding low-frequency semantics necessitates reducing high-frequency components, creating tension between semantic encoding and high-frequency decoding. To resolve this challenge, we propose a new \textbf{\color{ddt}D}ecoupled \textbf{\color{ddt}D}iffusion \textbf{\color{ddt}T}ransformer~(\textbf{\color{ddt}DDT}), with a decoupled design of a dedicated condition encoder for semantic extraction alongside a specialized velocity decoder. Our experiments reveal that a more substantial encoder yields performance improvements as model size increases. For ImageNet $256\times256$, Our DDT-XL/2 achieves a new state-of-the-art performance of {1.31 FID}~(nearly $4\times$ faster training convergence compared to previous diffusion transformers). For ImageNet $512\times512$, Our DDT-XL/2 achieves a new state-of-the-art FID of 1.28. Additionally, as a beneficial by-product, our decoupled architecture enhances inference speed by enabling the sharing self-condition between adjacent denoising steps. To minimize performance degradation, we propose a novel statistical dynamic programming approach to identify optimal sharing strategies.

cs.CV↗

arXiv2025

Tra-MoE: Learning Trajectory Prediction Model from Multiple Domains for Adaptive Policy Conditioning

Learning from multiple domains is a primary factor that influences the generalization of a single unified robot system. In this paper, we aim to learn the trajectory prediction model by using broad out-of-domain data to improve its performance and generalization ability. Trajectory model is designed to predict any-point trajectories in the current frame given an instruction and can provide detailed control guidance for robotic policy learning. To handle the diverse out-of-domain data distribution, we propose a sparsely-gated MoE (\textbf{Top-1} gating strategy) architecture for trajectory model, coined as \textbf{Tra-MoE}. The sparse activation design enables good balance between parameter cooperation and specialization, effectively benefiting from large-scale out-of-domain data while maintaining constant FLOPs per token. In addition, we further introduce an adaptive policy conditioning technique by learning 2D mask representations for predicted trajectories, which is explicitly aligned with image observations to guide action prediction more flexibly. We perform extensive experiments on both simulation and real-world scenarios to verify the effectiveness of Tra-MoE and adaptive policy conditioning technique. We also conduct a comprehensive empirical study to train Tra-MoE, demonstrating that our Tra-MoE consistently exhibits superior performance compared to the dense baseline model, even when the latter is scaled to match Tra-MoE's parameter count.

cs.RO↗

arXiv2025

LeviTor: 3D Trajectory Oriented Image-to-Video Synthesis

The intuitive nature of drag-based interaction has led to its growing adoption for controlling object trajectories in image-to-video synthesis. Still, existing methods that perform dragging in the 2D space usually face ambiguity when handling out-of-plane movements. In this work, we augment the interaction with a new dimension, i.e., the depth dimension, such that users are allowed to assign a relative depth for each point on the trajectory. That way, our new interaction paradigm not only inherits the convenience from 2D dragging, but facilitates trajectory control in the 3D space, broadening the scope of creativity. We propose a pioneering method for 3D trajectory control in image-to-video synthesis by abstracting object masks into a few cluster points. These points, accompanied by the depth information and the instance information, are finally fed into a video diffusion model as the control signal. Extensive experiments validate the effectiveness of our approach, dubbed LeviTor, in precisely manipulating the object movements when producing photo-realistic videos from static images. Our code is available at: https://github.com/ant-research/LeviTor.

cs.CV↗

arXiv2025

Multiple Object Tracking as ID Prediction

Multi-Object Tracking (MOT) has been a long-standing challenge in video understanding. A natural and intuitive approach is to split this task into two parts: object detection and association. Most mainstream methods employ meticulously crafted heuristic techniques to maintain trajectory information and compute cost matrices for object matching. Although these methods can achieve notable tracking performance, they often require a series of elaborate handcrafted modifications while facing complicated scenarios. We believe that manually assumed priors limit the method's adaptability and flexibility in learning optimal tracking capabilities from domain-specific data. Therefore, we introduce a new perspective that treats Multiple Object Tracking as an in-context ID Prediction task, transforming the aforementioned object association into an end-to-end trainable task. Based on this, we propose a simple yet effective method termed MOTIP. Given a set of trajectories carried with ID information, MOTIP directly decodes the ID labels for current detections to accomplish the association process. Without using tailored or sophisticated architectures, our method achieves state-of-the-art results across multiple benchmarks by solely leveraging object-level features as tracking cues. The simplicity and impressive results of MOTIP leave substantial room for future advancements, thereby making it a promising baseline for subsequent research. Our code and checkpoints are released at https://github.com/MCG-NJU/MOTIP.

cs.CV↗

arXiv2025

MiLA: Multi-view Intensive-fidelity Long-term Video Generation World Model for Autonomous Driving

In recent years, data-driven techniques have greatly advanced autonomous driving systems, but the need for rare and diverse training data remains a challenge, requiring significant investment in equipment and labor. World models, which predict and generate future environmental states, offer a promising solution by synthesizing annotated video data for training. However, existing methods struggle to generate long, consistent videos without accumulating errors, especially in dynamic scenes. To address this, we propose MiLA, a novel framework for generating high-fidelity, long-duration videos up to one minute. MiLA utilizes a Coarse-to-Re(fine) approach to both stabilize video generation and correct distortion of dynamic objects. Additionally, we introduce a Temporal Progressive Denoising Scheduler and Joint Denoising and Correcting Flow modules to improve the quality of generated videos. Extensive experiments on the nuScenes dataset show that MiLA achieves state-of-the-art performance in video generation quality. For more information, visit the project website: https://github.com/xiaomi-mlab/mila.github.io.

cs.CV↗

arXiv2025

CaReBench: A Fine-Grained Benchmark for Video Captioning and Retrieval

Video understanding, including video captioning and retrieval, is still a great challenge for video-language models (VLMs). The existing video retrieval and caption benchmarks only include short descriptions, limits their ability of detailed video understanding evaluation. To address this problem, we present CaReBench, a testing benchmark for fine-grained video captioning and retrieval with 1,000 high-quality pairs of videos and human-annotated detailed captions. Uniquely, it provides manually separated spatial annotations and temporal annotations for each video. Based on this design, we introduce two evaluation metrics, ReBias and CapST, specifically tailored for video retrieval and video captioning tasks, respectively. These metrics enable a comprehensive investigation into the spatial and temporal biases inherent in VLMs. In addition, to handle both video retrieval and video captioning tasks in a unified framework, we develop a simple baseline based on a Multimodal Language Model (MLLM). By implementing a two-stage Supervised Fine-Tuning (SFT), we fully unlock the potential of MLLM, enabling it not only to generate detailed video descriptions but also to extract video features. Surprisingly, experimental results demonstrate that, compared to the CLIP-based models designed for retrieval and the popular MLLMs skilled in video captioning, our baseline shows competitive performance in both fine-grained video retrieval and video detailed captioning.

cs.CV↗

arXiv2025

Make Your Training Flexible: Towards Deployment-Efficient Video Models

Popular video training methods mainly operate on a fixed number of tokens sampled from a predetermined spatiotemporal grid, resulting in sub-optimal accuracy-computation trade-offs due to inherent video redundancy. They also lack adaptability to varying computational budgets for downstream tasks, hindering applications of the most competitive model in real-world scenes. We thus propose a new test setting, Token Optimization, for maximized input information across budgets, which optimizes the size-limited set of input tokens through token selection from more suitably sampled videos. To this end, we propose a novel augmentation tool termed Flux. By making the sampling grid flexible and leveraging token selection, it is easily adopted in most popular video training frameworks, boosting model robustness with nearly no additional cost. We integrate Flux in large-scale video pre-training, and the resulting FluxViT establishes new state-of-the-art results across extensive tasks at standard costs. Notably, with 1/4 tokens only, it can still match the performance of previous state-of-the-art models with Token Optimization, yielding nearly 90\% savings. All models and data are available at https://github.com/OpenGVLab/FluxViT.

cs.CV↗

arXiv2025

VideoMAP: Toward Scalable Mamba-based Video Autoregressive Pretraining

Recent Mamba-based architectures for video understanding demonstrate promising computational efficiency and competitive performance, yet struggle with overfitting issues that hinder their scalability. To overcome this challenge, we introduce VideoMAP, a Hybrid Mamba-Transformer framework featuring a novel pre-training approach. VideoMAP uses a 4:1 Mamba-to-Transformer ratio, effectively balancing computational cost and model capacity. This architecture, combined with our proposed frame-wise masked autoregressive pre-training strategy, delivers significant performance gains when scaling to larger models. Additionally, VideoMAP exhibits impressive sample efficiency, significantly outperforming existing methods with less training data. Experiments show that VideoMAP outperforms existing models across various datasets, including Kinetics-400, Something-Something V2, Breakfast, and COIN. Furthermore, we demonstrate the potential of VideoMAP as a visual encoder for multimodal large language models, highlighting its ability to reduce memory usage and enable the processing of longer video sequences. The code is open-source at https://github.com/yunzeliu/MAP

cs.CV↗

arXiv2025

Tuning Electrode Wettability to Optimize Nanobubble Nucleation and Reaction Rates in Electrochemical Gas-Evolving Reactions

Bubble formation in electrochemical system often hinders reaction efficiency by reducing active surface area and obstructing mass transfer, yet the mechanisms governing their nanoscale nucleation dynamics and impact remains unclear. In this study, we used molecular dynamics simulations to explore nanobubble nucleation and reaction rates during water electrolysis on planar- and nano-electrodes, with systematically tuning electrode wettability through water-electrode and gas-electrode interactions. We identified distinct nucleation regimes: gas layers, surface nanobubbles, bulk nanobubbles, and no nanobubbles, and revealed a volcano-shaped relationship between wettability and reaction rate, where optimal wettability strikes a balance between suppressing bubbles and ensuring sufficient reactant availability to maximize performance. Nanoelectrodes consistently exhibit higher current densities compared to planar electrodes with the same wettability, due to pronounced edge effects. Furthermore, moderate driving forces enhance reaction rates without triggering surface bubble formation, while excessive driving forces induce surface nanobubble nucleation, leading to suppressed reaction rates and complex dynamics driven by bubble growth and detachment. These findings highlight the importance of fine-tuning wettability and reaction driving forces to optimize gas-evolving electrochemical systems at the nanoscale and underscore the need for multiscale simulation frameworks integrating atomic-scale reaction kinetics, nanoscale bubble nucleation, and microscale bubble dynamics to fully understand bubble behavior and its impact on performance.

cond-mat.soft↗

arXiv2025

A Large-Scale Study on Video Action Dataset Condensation

Recently, dataset condensation has made significant progress in the image domain. Unlike images, videos possess an additional temporal dimension, which harbors considerable redundant information, making condensation even more crucial. However, video dataset condensation still remains an underexplored area. We aim to bridge this gap by providing a large-scale study with systematic design and fair comparison. Specifically, our work delves into three key aspects to provide valuable empirical insights: (1) temporal processing of video data, (2) the evaluation protocol for video dataset condensation, and (3) adaptation of condensation algorithms to the space-time domain. From this study, we derive several intriguing observations: (i) labeling methods greatly influence condensation performance, (ii) simple sliding-window sampling is effective for temporal processing, and (iii) dataset distillation methods perform better in challenging scenarios, while sample selection methods excel in easier ones. Furthermore, we propose a unified evaluation protocol for the fair comparison of different condensation algorithms and achieve state-of-the-art results on four widely-used action recognition datasets: HMDB51, UCF101, SSv2 and K400. Our code is available at https://github.com/MCG-NJU/Video-DC.

cs.CV↗