arXiv subjects

Song Bai

Publications and source records attributed to Song Bai.

At least 19 recordsLinked to original sources

arXiv2026

Report of the 5th PVUW Challenge: Towards More Diverse Modalities in Pixel-Level Understanding

This report summarizes the objectives, datasets, and top-performing methodologies of the 2026 Pixel-level Video Understanding in the Wild (PVUW) Challenge, hosted at CVPR 2026, which evaluates state-of-the-art models under highly unconstrained conditions. To provide a comprehensive assessment, the 2026 edition features three specialized tracks: the MOSE track for tracking objects within densely cluttered and severely occluded scenarios; the MeViS-Text track for localizing targets via motion-focused linguistic expressions; and the newly inaugurated MeViS-Audio track, which pioneers acoustic-driven object segmentation. By introducing previously unreleased challenging data and analyzing the cutting-edge, multimodal solutions submitted by participants, this report highlights the community's latest technical advancements and charts promising future directions for robust video scene comprehension.

cs.CV↗

arXiv2026

MiroThinker: Pushing the Performance Boundaries of Open-Source Research Agents via Model, Context, and Interactive Scaling

We present MiroThinker v1.0, an open-source research agent designed to advance tool-augmented reasoning and information-seeking capabilities. Unlike previous agents that only scale up model size or context length, MiroThinker explores interaction scaling at the model level, systematically training the model to handle deeper and more frequent agent-environment interactions as a third dimension of performance improvement. Unlike LLM test-time scaling, which operates in isolation and risks degradation with longer reasoning chains, interactive scaling leverages environment feedback and external information acquisition to correct errors and refine trajectories. Through reinforcement learning, the model achieves efficient interaction scaling: with a 256K context window, it can perform up to 600 tool calls per task, enabling sustained multi-turn reasoning and complex real-world research workflows. Across four representative benchmarks-GAIA, HLE, BrowseComp, and BrowseComp-ZH-the 72B variant achieves up to 81.9%, 37.7%, 47.1%, and 55.6% accuracy respectively, surpassing previous open-source agents and approaching commercial counterparts such as GPT-5-high. Our analysis reveals that MiroThinker benefits from interactive scaling consistently: research performance improves predictably as the model engages in deeper and more frequent agent-environment interactions, demonstrating that interaction depth exhibits scaling behaviors analogous to model size and context length. These findings establish interaction scaling as a third critical dimension for building next-generation open research agents, complementing model capacity and context windows.

cs.CL↗

arXiv2026

CodeDance: A Dynamic Tool-integrated MLLM for Executable Visual Reasoning

Recent releases such as o3 highlight human-like "thinking with images" reasoning that combines tool use with stepwise verification, yet most open-source approaches still rely on text-only chains, rigid visual schemas, or single-step pipelines, limiting flexibility, interpretability, and transferability on complex tasks. We introduce CodeDance, which explores executable code as a general solver for visual reasoning. Unlike fixed-schema calls (e.g., only predicting bounding-box coordinates), CodeDance defines, composes, and executes code to orchestrate multiple tools, compute intermediate results, and render visual artifacts (e.g., boxes, lines, plots) that support transparent, self-checkable reasoning. To guide this process, we introduce a reward for balanced and adaptive tool calling, which balances exploration with efficiency and mitigates tool overuse. Interestingly, beyond the expected capabilities taught by atomic supervision, we empirically observe novel emergent behaviors during RL training: CodeDance demonstrates novel tool invocations, unseen compositions, and cross-task transfer. These behaviors arise without task-specific fine-tuning, suggesting a general and scalable mechanism for executable visual reasoning. Extensive experiments across reasoning benchmarks (e.g., visual search, math, chart QA) show that CodeDance not only consistently outperforms schema-driven and text-only baselines, but also surpasses closed models such as GPT-4o and larger open-source models.

cs.CV↗

arXiv2026

Monocular Normal Estimation via Shading Sequence Estimation

Monocular normal estimation aims to estimate the normal map from a single RGB image of an object under arbitrary lights. Existing methods rely on deep models to directly predict normal maps. However, they often suffer from 3D misalignment: while the estimated normal maps may appear to have a correct appearance, the reconstructed surfaces often fail to align with the geometric details. We argue that this misalignment stems from the current paradigm: the model struggles to distinguish and reconstruct varying geometry represented in normal maps, as the differences in underlying geometry are reflected only through relatively subtle color variations. To address this issue, we propose a new paradigm that reformulates normal estimation as shading sequence estimation, where shading sequences are more sensitive to various geometric information. Building on this paradigm, we present RoSE, a method that leverages image-to-video generative models to predict shading sequences. The predicted shading sequences are then converted into normal maps by solving a simple ordinary least-squares problem. To enhance robustness and better handle complex objects, RoSE is trained on a synthetic dataset, MultiShade, with diverse shapes, materials, and light conditions. Experiments demonstrate that RoSE achieves state-of-the-art performance on real-world benchmark datasets for object-based monocular normal estimation.

cs.CV↗

arXiv2025

MOSEv2: A More Challenging Dataset for Video Object Segmentation in Complex Scenes

Video object segmentation (VOS) aims to segment specified target objects throughout a video. Although state-of-the-art methods have achieved impressive performance (e.g., 90+% J&F) on benchmarks such as DAVIS and YouTube-VOS, these datasets primarily contain salient, dominant, and isolated objects, limiting their generalization to real-world scenarios. To bridge this gap, the coMplex video Object SEgmentation (MOSEv1) dataset was introduced to facilitate VOS research in complex scenes. Building on the foundations and insights of MOSEv1, we present MOSEv2, a significantly more challenging dataset designed to further advance VOS methods under real-world conditions. MOSEv2 consists of 5,024 videos and 701,976 high-quality masks for 10,074 objects across 200 categories. Compared to its predecessor, MOSEv2 introduces much greater scene complexity, including {more frequent object disappearance and reappearance, severe occlusions and crowding, smaller objects, as well as a range of new challenges such as adverse weather (e.g., rain, snow, fog), low-light scenes (e.g., nighttime, underwater), multi-shot sequences, camouflaged objects, non-physical targets (e.g., shadows, reflections), and scenarios requiring external knowledge.} We benchmark 20 representative VOS methods under 5 different settings and observe consistent performance drops on MOSEv2. For example, SAM2 drops from 76.4% on MOSEv1 to only 50.9% on MOSEv2. We further evaluate 9 video object tracking methods and observe similar declines, demonstrating that MOSEv2 poses challenges across tasks. These results highlight that despite strong performance on existing datasets, current VOS methods still fall short under real-world complexities. Based on our analysis of the observed challenges, we further propose several practical tricks that enhance model performance. MOSEv2 is publicly available at https://MOSE.video.

cs.CV↗

arXiv2025

Text4Seg++: Advancing Image Segmentation via Generative Language Modeling

Multimodal Large Language Models (MLLMs) have shown exceptional capabilities in vision-language tasks. However, effectively integrating image segmentation into these models remains a significant challenge. In this work, we propose a novel text-as-mask paradigm that casts image segmentation as a text generation problem, eliminating the need for additional decoders and significantly simplifying the segmentation process. Our key innovation is semantic descriptors, a new textual representation of segmentation masks where each image patch is mapped to its corresponding text label. We first introduce image-wise semantic descriptors, a patch-aligned textual representation of segmentation masks that integrates naturally into the language modeling pipeline. To enhance efficiency, we introduce the Row-wise Run-Length Encoding (R-RLE), which compresses redundant text sequences, reducing the length of semantic descriptors by 74% and accelerating inference by $3\times$, without compromising performance. Building upon this, our initial framework Text4Seg achieves strong segmentation performance across a wide range of vision tasks. To further improve granularity and compactness, we propose box-wise semantic descriptors, which localizes regions of interest using bounding boxes and represents region masks via structured mask tokens called semantic bricks. This leads to our refined model, Text4Seg++, which formulates segmentation as a next-brick prediction task, combining precision, scalability, and generative efficiency. Comprehensive experiments on natural and remote sensing datasets show that Text4Seg++ consistently outperforms state-of-the-art models across diverse benchmarks without any task-specific fine-tuning, while remaining compatible with existing MLLM backbones. Our work highlights the effectiveness, scalability, and generalizability of text-driven image segmentation within the MLLM framework.

cs.CV↗

arXiv2025

Versatile Transition Generation with Image-to-Video Diffusion

Leveraging text, images, structure maps, or motion trajectories as conditional guidance, diffusion models have achieved great success in automated and high-quality video generation. However, generating smooth and rational transition videos given the first and last video frames as well as descriptive text prompts is far underexplored. We present VTG, a Versatile Transition video Generation framework that can generate smooth, high-fidelity, and semantically coherent video transitions. VTG introduces interpolation-based initialization that helps preserve object identity and handle abrupt content changes effectively. In addition, it incorporates dual-directional motion fine-tuning and representation alignment regularization to mitigate the limitations of pre-trained image-to-video diffusion models in motion smoothness and generation fidelity, respectively. To evaluate VTG and facilitate future studies on unified transition generation, we collected TransitBench, a comprehensive benchmark for transition generation covering two representative transition tasks: concept blending and scene transition. Extensive experiments show that VTG achieves superior transition performance consistently across all four tasks.

cs.CV↗

arXiv2025

TimeExpert: An Expert-Guided Video LLM for Video Temporal Grounding

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.

cs.CV↗

arXiv2025

Describe, Adapt and Combine: Empowering CLIP Encoders for Open-set 3D Object Retrieval

Open-set 3D object retrieval (3DOR) is an emerging task aiming to retrieve 3D objects of unseen categories beyond the training set. Existing methods typically utilize all modalities (i.e., voxels, point clouds, multi-view images) and train specific backbones before fusion. However, they still struggle to produce generalized representations due to insufficient 3D training data. Being contrastively pre-trained on web-scale image-text pairs, CLIP inherently produces generalized representations for a wide range of downstream tasks. Building upon it, we present a simple yet effective framework named Describe, Adapt and Combine (DAC) by taking only multi-view images for open-set 3DOR. DAC innovatively synergizes a CLIP model with a multi-modal large language model (MLLM) to learn generalized 3D representations, where the MLLM is used for dual purposes. First, it describes the seen category information to align with CLIP's training objective for adaptation during training. Second, it provides external hints about unknown objects complementary to visual cues during inference. To improve the synergy, we introduce an Additive-Bias Low-Rank adaptation (AB-LoRA), which alleviates overfitting and further enhances the generalization to unseen categories. With only multi-view images, DAC significantly surpasses prior arts by an average of +10.01\% mAP on four open-set 3DOR datasets. Moreover, its generalization is also validated on image-based and cross-dataset setups. Code is available at https://github.com/wangzhichuan123/DAC.

cs.CV↗

arXiv2025

PVUW 2025 Challenge Report: Advances in Pixel-level Understanding of Complex Videos in the Wild

This report provides a comprehensive overview of the 4th Pixel-level Video Understanding in the Wild (PVUW) Challenge, held in conjunction with CVPR 2025. It summarizes the challenge outcomes, participating methodologies, and future research directions. The challenge features two tracks: MOSE, which focuses on complex scene video object segmentation, and MeViS, which targets motion-guided, language-based video segmentation. Both tracks introduce new, more challenging datasets designed to better reflect real-world scenarios. Through detailed evaluation and analysis, the challenge offers valuable insights into the current state-of-the-art and emerging trends in complex video segmentation. More information can be found on the workshop website: https://pvuw.github.io/.

cs.CV↗

arXiv2025

Liquid: Language Models are Scalable and Unified Multi-modal Generators

We present Liquid, an auto-regressive generation paradigm that seamlessly integrates visual comprehension and generation by tokenizing images into discrete codes and learning these code embeddings alongside text tokens within a shared feature space for both vision and language. Unlike previous multimodal large language model (MLLM), Liquid achieves this integration using a single large language model (LLM), eliminating the need for external pretrained visual embeddings such as CLIP. For the first time, Liquid uncovers a scaling law that performance drop unavoidably brought by the unified training of visual and language tasks diminishes as the model size increases. Furthermore, the unified token space enables visual generation and comprehension tasks to mutually enhance each other, effectively removing the typical interference seen in earlier models. We show that existing LLMs can serve as strong foundations for Liquid, saving 100x in training costs while outperforming Chameleon in multimodal capabilities and maintaining language performance comparable to mainstream LLMs like LLAMA2. Liquid also outperforms models like SD v2.1 and SD-XL (FID of 5.47 on MJHQ-30K), excelling in both vision-language and text-only tasks. This work demonstrates that LLMs such as Qwen2.5 and GEMMA2 are powerful multimodal generators, offering a scalable solution for enhancing both vision-language understanding and generation. The code and models will be released at https://github.com/FoundationVision/Liquid.

cs.CV↗

arXiv2025

Learning to Animate Images from A Few Videos to Portray Delicate Human Actions

Despite recent progress, video generative models still struggle to animate static images into videos that portray delicate human actions, particularly when handling uncommon or novel actions whose training data are limited. In this paper, we explore the task of learning to animate images to portray delicate human actions using a small number of videos -- 16 or fewer -- which is highly valuable for real-world applications like video and movie production. Learning generalizable motion patterns that smoothly transition from user-provided reference images in a few-shot setting is highly challenging. We propose FLASH (Few-shot Learning to Animate and Steer Humans), which learns generalizable motion patterns by forcing the model to reconstruct a video using the motion features and cross-frame correspondences of another video with the same motion but different appearance. This encourages transferable motion learning and mitigates overfitting to limited training data. Additionally, FLASH extends the decoder with additional layers to propagate details from the reference image to generated frames, improving transition smoothness. Human judges overwhelmingly favor FLASH, with 65.78\% of 488 responses prefer FLASH over baselines. We strongly recommend watching the videos in the website: https://lihaoxin05.github.io/human_action_animation/, as motion artifacts are hard to notice from images.

cs.CV↗

arXiv2024

PartGLEE: A Foundation Model for Recognizing and Parsing Any Objects

We present PartGLEE, a part-level foundation model for locating and identifying both objects and parts in images. Through a unified framework, PartGLEE accomplishes detection, segmentation, and grounding of instances at any granularity in the open world scenario. Specifically, we propose a Q-Former to construct the hierarchical relationship between objects and parts, parsing every object into corresponding semantic parts. By incorporating a large amount of object-level data, the hierarchical relationships can be extended, enabling PartGLEE to recognize a rich variety of parts. We conduct comprehensive studies to validate the effectiveness of our method, PartGLEE achieves the state-of-the-art performance across various part-level tasks and obtain competitive results on object-level tasks. The proposed PartGLEE significantly enhances hierarchical modeling capabilities and part-level perception over our previous GLEE model. Further analysis indicates that the hierarchical cognitive ability of PartGLEE is able to facilitate a detailed comprehension in images for mLLMs. The model and code will be released at https://provencestar.github.io/PartGLEE-Vision/ .

cs.CV↗

arXiv2024

PVUW 2024 Challenge on Complex Video Understanding: Methods and Results

Pixel-level Video Understanding in the Wild Challenge (PVUW) focus on complex video understanding. In this CVPR 2024 workshop, we add two new tracks, Complex Video Object Segmentation Track based on MOSE dataset and Motion Expression guided Video Segmentation track based on MeViS dataset. In the two new tracks, we provide additional videos and annotations that feature challenging elements, such as the disappearance and reappearance of objects, inconspicuous small objects, heavy occlusions, and crowded environments in MOSE. Moreover, we provide a new motion expression guided video segmentation dataset MeViS to study the natural language-guided video understanding in complex environments. These new videos, sentences, and annotations enable us to foster the development of a more comprehensive and robust pixel-level understanding of video scenes in complex environments and realistic scenarios. The MOSE challenge had 140 registered teams in total, 65 teams participated the validation phase and 12 teams made valid submissions in the final challenge phase. The MeViS challenge had 225 registered teams in total, 50 teams participated the validation phase and 5 teams made valid submissions in the final challenge phase.

cs.CV↗

arXiv2024

DIRECT-3D: Learning Direct Text-to-3D Generation on Massive Noisy 3D Data

We present DIRECT-3D, a diffusion-based 3D generative model for creating high-quality 3D assets (represented by Neural Radiance Fields) from text prompts. Unlike recent 3D generative models that rely on clean and well-aligned 3D data, limiting them to single or few-class generation, our model is directly trained on extensive noisy and unaligned `in-the-wild' 3D assets, mitigating the key challenge (i.e., data scarcity) in large-scale 3D generation. In particular, DIRECT-3D is a tri-plane diffusion model that integrates two innovations: 1) A novel learning framework where noisy data are filtered and aligned automatically during the training process. Specifically, after an initial warm-up phase using a small set of clean data, an iterative optimization is introduced in the diffusion process to explicitly estimate the 3D pose of objects and select beneficial data based on conditional density. 2) An efficient 3D representation that is achieved by disentangling object geometry and color features with two separate conditional diffusion models that are optimized hierarchically. Given a prompt input, our model generates high-quality, high-resolution, realistic, and complex 3D objects with accurate geometric details in seconds. We achieve state-of-the-art performance in both single-class generation and text-to-3D generation. We also demonstrate that DIRECT-3D can serve as a useful 3D geometric prior of objects, for example to alleviate the well-known Janus problem in 2D-lifting methods such as DreamFusion. The code and models are available for research purposes at: https://github.com/qihao067/direct3d.

cs.CV↗

arXiv2024

Mimicking the Oracle: An Initial Phase Decorrelation Approach for Class Incremental Learning

Class Incremental Learning (CIL) aims at learning a multi-class classifier in a phase-by-phase manner, in which only data of a subset of the classes are provided at each phase. Previous works mainly focus on mitigating forgetting in phases after the initial one. However, we find that improving CIL at its initial phase is also a promising direction. Specifically, we experimentally show that directly encouraging CIL Learner at the initial phase to output similar representations as the model jointly trained on all classes can greatly boost the CIL performance. Motivated by this, we study the difference between a naïvely-trained initial-phase model and the oracle model. Specifically, since one major difference between these two models is the number of training classes, we investigate how such difference affects the model representations. We find that, with fewer training classes, the data representations of each class lie in a long and narrow region; with more training classes, the representations of each class scatter more uniformly. Inspired by this observation, we propose Class-wise Decorrelation (CwD) that effectively regularizes representations of each class to scatter more uniformly, thus mimicking the model jointly trained with all classes (i.e., the oracle model). Our CwD is simple to implement and easy to plug into existing methods. Extensive experiments on various benchmark datasets show that CwD consistently and significantly improves the performance of existing state-of-the-art methods by around 1\% to 3\%. Code will be released.

cs.CV↗

arXiv2024

Towards Understanding and Mitigating Dimensional Collapse in Heterogeneous Federated Learning

Federated learning aims to train models collaboratively across different clients without the sharing of data for privacy considerations. However, one major challenge for this learning paradigm is the {\em data heterogeneity} problem, which refers to the discrepancies between the local data distributions among various clients. To tackle this problem, we first study how data heterogeneity affects the representations of the globally aggregated models. Interestingly, we find that heterogeneous data results in the global model suffering from severe {\em dimensional collapse}, in which representations tend to reside in a lower-dimensional space instead of the ambient space. Moreover, we observe a similar phenomenon on models locally trained on each client and deduce that the dimensional collapse on the global model is inherited from local models. In addition, we theoretically analyze the gradient flow dynamics to shed light on how data heterogeneity result in dimensional collapse for local models. To remedy this problem caused by the data heterogeneity, we propose {\sc FedDecorr}, a novel method that can effectively mitigate dimensional collapse in federated learning. Specifically, {\sc FedDecorr} applies a regularization term during local training that encourages different dimensions of representations to be uncorrelated. {\sc FedDecorr}, which is implementation-friendly and computationally-efficient, yields consistent improvements over baselines on standard benchmark datasets. Code: https://github.com/bytedance/FedDecorr.

cs.LG↗

arXiv2024

DragDiffusion: Harnessing Diffusion Models for Interactive Point-based Image Editing

Accurate and controllable image editing is a challenging task that has attracted significant attention recently. Notably, DragGAN is an interactive point-based image editing framework that achieves impressive editing results with pixel-level precision. However, due to its reliance on generative adversarial networks (GANs), its generality is limited by the capacity of pretrained GAN models. In this work, we extend this editing framework to diffusion models and propose a novel approach DragDiffusion. By harnessing large-scale pretrained diffusion models, we greatly enhance the applicability of interactive point-based editing on both real and diffusion-generated images. Our approach involves optimizing the diffusion latents to achieve precise spatial control. The supervision signal of this optimization process is from the diffusion model's UNet features, which are known to contain rich semantic and geometric information. Moreover, we introduce two additional techniques, namely LoRA fine-tuning and latent-MasaCtrl, to further preserve the identity of the original image. Lastly, we present a challenging benchmark dataset called DragBench -- the first benchmark to evaluate the performance of interactive point-based image editing methods. Experiments across a wide range of challenging cases (e.g., images with multiple objects, diverse object categories, various styles, etc.) demonstrate the versatility and generality of DragDiffusion. Code: https://github.com/Yujun-Shi/DragDiffusion.

cs.CV↗