CVPR 2026 论文列表

Accurate robot segmentation is a fundamental capability for robotic perception. It enables precise visual servoing for VLA systems, scalable robot-centric data augmentation, accurate real-to-sim transfer, and reliable safety monitoring in dynamic human-robot environments. Despite the strong capabilities of modern segmentation models, surprisingly it remains challenging to segment robots. This is due to robot embodiment diversity, appearance ambiguity, structural complexity, and rapid shape changes. Embracing these challenges, we introduce RobotSeg, a foundation model for robot segmentation in image and video. RobotSeg is built upon the versatile SAM 2 foundation model but addresses its three limitations for robot segmentation, namely the lack of adaptation to articulated robots, reliance on manual prompts, and the need for per-frame training mask annotations, by introducing a structure-enhanced memory associator, a robot prompt generator, and a label-efficient training strategy. These innovations collectively enable a structure-aware, automatic, and label-efficient solution. We further construct the video robot segmentation (VRS) dataset comprising over 2.8k videos (138k frames) with diverse robot embodiments and environments. Extensive experiments demonstrate that RobotSeg achieves state-of-the-art performance on both images and videos, establishing a strong foundation for future advances in robot perception.

This paper focuses on a highly practical scenario: how to continue benefiting from the advantages of multi-modal image fusion under harsh conditions when only visible imaging sensors are available. To achieve this goal, we propose a novel concept of single image fusion, which extends conventional data-level fusion to the knowledge level. Specifically, we develop MagicFuse, a novel single image fusion framework capable of deriving a comprehensive cross-spectral scene representation from a single low-quality visible image. MagicFuse first introduces an intra-spectral knowledge reinforcement branch and a cross-spectral knowledge generation branch based on the diffusion models. They mine scene information obscured in the visible spectrum and learn thermal radiation distribution patterns transferred to the infrared spectrum, respectively. Building on them, we design a multi-domain knowledge fusion branch that integrates the probabilistic noise from the diffusion streams of these two branches, from which a cross-spectral scene representation can be obtained through successive sampling. Then, we impose both visual and semantic constraints to ensure that this scene representation can satisfy human observation while supporting downstream semantic decision-making. Extensive experiments show that our MagicFuse achieves visual and semantic representation performance comparable to or even better than state-of-the-art fusion methods with multi-modal inputs, despite relying solely on a single degraded visible image. The code is publicly available at https://github.com/zhayanping/MagicFuse.

The advent of one-step text-to-image (T2I) models offers unprecedented synthesis speed. However, their application to text-guided image editing remains severely hampered, as forcing existing training-free editors into a single inference step fails. This failure manifests as severe object distortion and a critical loss of consistency in non-edited regions, resulting from the high-energy, erratic trajectories produced by naive vector arithmetic on the models' structured fields. To address this problem, we introduce ChordEdit, a model agnostic, training-free, and inversion-free method that facilitates high-fidelity one-step editing. We recast editing as a transport problem between the source and target distributions defined by the source and target text prompts. Leveraging dynamic optimal transport theory, we derive a principled, low-energy control strategy. This strategy yields a smoothed, variance-reduced editing field that is inherently stable, facilitating the field to be traversed in a single, large integration step. A theoretically grounded and experimentally validated approach allows ChordEdit to deliver fast, lightweight and precise edits, finally achieving true real-time editing on these challenging models.

Although text-to-image diffusion models exhibit remarkable generative power, concept erasure techniques are essential for their safe deployment to prevent the creation of harmful content.This has fostered a dynamic interplay between the development of erasure defenses and the adversarial probes designed to bypass them, and this co-evolution has progressively enhanced the efficacy of erasure methods.However, this adversarial co-evolution has converged on a narrow, text-centric paradigm that equates erasure with severing the text-to-image mapping, ignoring that the underlying visual knowledge related to undesired concepts still persist.To substantiate this claim, we investigate from a visual perspective, leveraging DDIM inversion to probe whether a generative pathway for the erased concept can still be found.However, identifying such a visual generative pathway is challenging because standard text-guided DDIM inversion is actively resisted by text-centric defenses within the erased model.To address this, we introduce TINA, a novel Text-free INversion Attack, which enforces this visual-only probe by operating under a null-text condition, thereby avoiding existing text-centric defenses.Moreover, TINA integrates an optimization procedure to overcome the accumulating approximation errors that arise when standard inversion operates without its usual textual guidance.Our experiments demonstrate that TINA successfully regenerates erased concepts from models treated with state-of-the-art unlearning.The success of TINA proves that current methods merely obscure concepts, highlighting an urgent need for paradigms that operate directly on internal visual knowledge.Code will be released upon acceptance.

Recent unified models have made unprecedented progress in both understanding and generation. However, while most of them accept multi-modal inputs, they typically produce only single-modality outputs. This challenge of producing interleaved content is mainly due to training data scarcity and the difficulty of modeling long-range cross-modal context. To address this issue, we decompose interleaved generation into textual planning and visual consistency modeling, and introduce a framework consisting of a planner and a visualizer. The planner produces dense textual descriptions for visual content, while the visualizer synthesizes images accordingly. Under this guidance, we construct large-scale textual-proxy interleaved data (where visual content is represented in text) to train the planner, and curate reference-guided image data to train the visualizer. These designs give rise to Wan-Weaver, which exhibits emergent interleaved generation ability with long-range textual coherence and visual consistency. Meanwhile, the integration of diverse understanding and generation data into planner training enables Wan-Weaver to achieve robust task reasoning and generation proficiency. To assess the model's capability in interleaved generation, we further construct a benchmark that spans a wide range of use cases across multiple dimensions. Extensive experiments demonstrate that, even without access to any real interleaved data, Wan-Weaver achieves superior performance over existing methods.

Efficient image compression relies on modeling both local and global redundancy. Most state-of-the-art (SOTA) learned image compression (LIC) methods are based on CNNs or Transformers, which are inherently rigid. Standard CNN kernels and window-based attention mechanisms impose fixed receptive fields and static connectivity patterns, which potentially couple non-redundant pixels simply due to their proximity in Euclidean space. This rigidity limits the model's ability to adaptively capture spatially varying redundancy across the image, particularly at the global level. To overcome these limitations, we propose a content-adaptive image compression framework based on Graph Neural Networks (GNNs). Specifically, our approach constructs dual-scale graphs that enable flexible, data-driven receptive fields. Furthermore, we introduce adaptive connectivity by dynamically adjusting the number of neighbors for each node based on local content complexity. These innovations empower our Graph-based Learned Image Compression (GLIC) model to effectively model diverse redundancy patterns across images, leading to more efficient and adaptive compression. Experiments demonstrate that GLIC achieves state-of-the-art performance, achieving BD-rate reductions of 19.29%, 21.69%, and 18.71% relative to VTM-9.1 on Kodak, Tecnick, and CLIC, respectively. Code will be released at https://github.com/UnoC-727/GLIC.

Composed Image Retrieval (CIR) aims to retrieve target images based on a hybrid query comprising a reference image and a modification text. Early dual-tower Vision-Language Models (VLMs) struggle with cross-modality compositional reasoning required for this task. While adapting generative Multimodal Large Language Models (MLLMs) for retrieval offers a promising direction, we identify that this strategy overlooks a fundamental issue: compressing a generative MLLM into a single-embedding discriminative retriever triggers a paradigm conflict, which leads to Capability Degradation--the deterioration of native fine-grained reasoning after retrieval adaptation. To address this challenge, we propose ReCALL, a model-agnostic framework that follows a diagnose-generate-refine pipeline: First, we diagnose cognitive blind spots of the retriever via self-guided informative instance mining. Next, we generate corrective instructions and triplets by prompting the foundation MLLM and conduct quality control with VQA-based consistency filtering. Finally, we refine the retriever through continual training on these triplets with a grouped contrastive scheme, thereby internalizing fine-grained visual-semantic distinctions and realigning the discriminative embedding space of retriever with intrinsic compositional reasoning within the MLLM. Extensive experiments on CIRR and FashionIQ show that ReCALL consistently recalibrates degraded capabilities and achieves state-of-the-art performance. Code is available at https://github.com/RemRico/Recall.

We introduce MoS (Mixture of States), a novel fusion paradigm for multimodal diffusion models that merges modalities using flexible, state-based interactions. The core of MoS is a learnable, token-wise router that creates denoising timestep- and input-dependent interactions between modalities' hidden states, precisely aligning token-level features with the diffusion trajectory. This router sparsely selects the top-k hidden states and is trained with an epsilon-greedy strategy, efficiently selecting contextual features with minimal learnable parameters and negligible computational overhead. We validate our design with text-to-image generation (MoS-Image) and editing (MoS-Editing), which achieve state-of-the-art results. With only 3B to 5B parameters, our models match or surpass counterparts up to 4xlarger. These findings establish MoS as a flexible and compute-efficient paradigm for scaling multimodal diffusion models. The project homepage is available at https://haozheliu-st.github.io/mos-homepage/

Generative models excel at motion synthesis for a fixed number of agents but struggle to generalize with variable agents. Based on limited, domain-specific data, existing methods employ autoregressive models to generate motion recursively, which suffer from inefficiency and error accumulation. We propose Unified Motion Flow (UMF), which consists of Pyramid Motion Flow (P-Flow) and Semi-Noise Motion Flow (S-Flow). UMF decomposes the number-free motion generation into a single-pass motion prior generation stage and multi-pass reaction generation stages. Specifically, UMF utilizes a unified latent space to bridge the distribution gap between heterogeneous motion datasets, enabling effective unified training. For motion prior generation, P-Flow operates on hierarchical resolutions conditioned on different noise levels, thereby mitigating computational overheads. For reaction generation, S-Flow learns a joint probabilistic path that adaptively performs reaction transformation and context reconstruction, alleviating error accumulation. Extensive results and user studies demonstrate UMF's effectiveness as a generalist model for multi-person motion generation from text. Project page: https://githubhgh.github.io/umf/.

Recent large vision-language models (LVLMs) have been applied to diverse VQA tasks. However, achieving practical performance typically requires task-specific fine-tuning with large numbers of image-text pairs, which are costly to collect. In this work, we study text-centric training, a setting where only textual descriptions are available and no real images are provided, as a paradigm for low-cost data scaling. Unlike images, whose collection is often restricted by privacy constraints and scarcity in niche domains, text is widely available. Moreover, text is easily editable, enabling automatic diversification and expansion with LLMs at minimal human effort. While this offers clear advantages over image collection in terms of scalability and cost, training on raw text without images still yields limited gains on VQA tasks because of the image-text modality gap. To address this issue, we propose a Text-Printed Image (TPI), which generates synthetic images by directly rendering the given textual description on a plain white canvas. This simple rendering projects text into the image modality and can be integrated into many existing LVLM training pipelines at low cost. Moreover, TPI preserves the semantics of the text, whereas text-to-image models often fail to do. Across four models and seven benchmarks, our systematic experiments show that TPI enables more effective text-centric training than synthetic images generated by a diffusion model. We further explore TPI as a low-cost data-augmentation strategy and demonstrate its practical utility. Overall, our findings highlight the significant potential of text-centric training and, more broadly, chart a path toward fully automated data generation for LVLMs.

Stereo image compression (SIC) has become increasingly vital with its applications surging in fields such as 3D reconstruction and autonomous navigation. Previous methods leverage cross-attention to model inter-view redundancy and employ autoregressive entropy models to predict probability distributions, achieving impressive rate-distortion performance. However, they suffer from slow coding speed due to the quadratic complexity of cross-attention mechanisms and the spatial autoregressive iterations of the entropy models. To address these limitations, we propose MambaSIC, which introduces two key innovations. First, we propose a Mamba-based stereo visual state space block (stereo VSSB) that leverages its linear complexity and long-range modeling capabilities to more rapidly and efficiently capture redundancy information between the two views. Second, to accelerate the compression process and enhance the accuracy of probability estimation, we introduce a bi-directional multi-reference entropy model that utilizes a checkerboard partitioning strategy and the stereo VSSB to get rich inter-view priors. Experimental results demonstrate that our MambaSIC outperforms the state-of-the-art methods in both compression performance and efficiency.

Diffusion Models have emerged as a leading class of generative models, yet their iterative sampling process remains computationally expensive. Timestep distillation is a promising technique to accelerate generation, but it often requires extensive training and leads to image quality degradation. Furthermore, fine-tuning these distilled models for specific objectives, such as aesthetic appeal or user preference, using Reinforcement Learning (RL) is notoriously unstable and easily falls into reward hacking. In this work, we introduce Flash-DMD, a novel framework that enables fast convergence with distillation and joint RL-based refinement.Specifically, we first propose an efficient timestep-aware distillation strategy that significantly reduces training cost with enhanced realism, outperforming DMD2 with only 2.1% its training cost. Second, we introduce a joint training scheme where the model is fine-tuned with an RL objective while the timestep distillation training continues simultaneously. We demonstrate that the stable, well-defined loss from the ongoing distillation acts as a powerful regularizer, effectively stabilizing the RL training process and preventing policy collapse. Extensive experiments on score-based and flow matching models show that our proposed Flash-DMD not only converges significantly faster but also achieves state-of-the-art generation quality in the few-step sampling regime, outperforming existing methods in visual quality, human preference, and text-image alignment metrics. Our work presents an effective paradigm for training efficient, high-fidelity, and stable generative models. Codes are attached in the supplementary.

Boundary representation (B-Rep) generation is a fundamental task in computer-aided design, yet the direct synthesis of high-fidelity and structurally valid B-Reps remains a major challenge. Existing deep generative methods suffer from two forms of brittleness: representation brittleness, caused by padding noise and feature contamination in the latent space, and generation brittleness, stemming from sequential error propagation and a train-inference mismatch due to non-differentiable validity enforcement. We propose HiFi-BRep, a novel framework that addresses these limitations through two synergistic contributions. First, a topology-aware encoder constructs a high-fidelity latent representation by eliminating padding via learnable queries and preventing feature contamination with topology-guided attention. Second, a single-stage decoder jointly predicts geometry and topology in parallel, embedding core manifold constraints as a differentiable learning objective. This design ensures mutual guidance between geometry and topology while avoiding cascaded errors. Extensive experiments show that HiFi-BRep significantly outperforms state-of-the-art methods in both structural validity and geometric fidelity, providing a robust solution for high-quality B-Rep synthesis. Code and models are publicly available at https://github.com/1nnoh/HiFi-BRep.

Human motions are compositional: complex behaviors can be described as combinations of simpler primitives. However, existing approaches primarily focus on forward modeling, e.g., learning holistic mappings from text to motion or composing a complex motion from a set of motion concepts. In this paper, we consider the inverse perspective: decomposing a holistic motion into semantically meaningful sub-components. We propose DeMoGen, a compositional training paradigm for decompositional learning that employs an energy-based diffusion model. This energy formulation directly captures the composed distribution of multiple motion concepts, enabling the model to discover them without relying on ground-truth motions for individual concepts. Within this paradigm, we introduce three training variants to encourage a decompositional understanding of motion: 1. DeMoGen-Exp explicitly trains on decomposed text prompts; 2. DeMoGen-OSS performs orthogonal self-supervised decomposition; 3. DeMoGen-SC enforces semantic consistency between original and decomposed text embeddings. These variants enable our approach to disentangle reusable motion primitives from complex motion sequences. We also demonstrate that the decomposed motion concepts can be flexibly recombined to generate diverse and novel motions, generalizing beyond the training distribution. Additionally, we construct a text-decomposed dataset to support compositional training, serving as an extended resource to facilitate text-to-motion generation and motion composition.

Existing literature typically treats various customized generation tasks (e.g., subject-customized generation, style-customized generation) as distinct and disjoint problems, with each task focusing solely on customizing a specific aspect of the reference image. However, we argue that the objectives of these different customization tasks are inherently complementary and can be mutually enhanced within a unified framework, as they fundamentally involve the disentanglement of multiple feature aspects from the reference image. To this end, we introduce USO, a Unified Simultaneous Optimization framework to simultaneously unify different customized tasks (i.e., subject and style). Specifically, USO introduces a cyclical data-model framework that connects these two tasks by a subject-for-style data curation pipeline and a style-for-subject model training pipeline. The subject-for-style data curation pipeline leverages a state-of-the-art subject-customized model to generate high-quality triplet data comprising content images, style images, and their corresponding stylized content images. Building on this foundation, the style-for-subject model training pipeline introduces an auxiliary style reward to simultaneously align style and content features, thereby reinforcing the model's ability to extract the desired style or content features from the reference image. Extensive experiments demonstrate that USO achieves state-of-the-art performance among open-source models, excelling in both subject consistency and style similarity. Code and model: https://github.com/bytedance/USO.

Style-conditioned scene text generation faces unique challenges in extracting precise text styles from complex backgrounds and maintaining fine-grained style consistency across characters, especially for multilingual scripts. We propose StyleTextGen, a novel framework that learns to perceive and replicate visual text styles across different languages and writing systems. Our approach features three key contributions: First, we introduce a dual-branch style encoder dedicated to style modeling, yielding robust multilingual text style representations in complex real-world scenes. Second, we design a text style consistency loss that enhances style coherence and improves overall visual quality. Third, we develop a mask-guided inference strategy that ensures precise style alignment between generated and reference text. To facilitate systematic evaluation, we construct StyleText-CE, a bilingual scene text style benchmark covering both monolingual and cross-lingual settings. Extensive experiments demonstrate that StyleTextGen significantly outperforms existing methods in style consistency and cross-lingual generalization, establishing new state-of-the-art performance in multilingual style-conditioned text generation.

Cross-modal biomedical signals such as pathology and genomics can provide richer and more robust semantic guidance for medical image representation learning. However, the availability of such guidance remains limited, as privacy constraints and acquisition costs severely restrict access to medical images paired with other biomedical data. A further challenge lies in modality discrepancy, which introduces intra-modal statistical bias and cross-modal noise, thereby degrading the quality of medical image representations. To address these challenges, we propose KAMP, a large language model (LLM)-driven multimodal pretraining framework for medical image representation learning. KAMP leverages textual priors as the semantic anchor to enhance medical image representations and align them with multimodal biomedical representations, enabling the learning of rich and robust features even when paired data are scarce. KAMP operates in three stages. First, the LLM generates personalized diagnostic knowledge from patient clinical text and imaging metadata. This knowledge is injected as a prior to enrich medical image representations and serves as a semantic anchor to reduce the representation gap between medical images and other biomedical modalities. Second, the LLM is optimized using Group Relative Policy Optimization (GRPO), with the cross-modal aligner pretrained in the first stage serving as the reward model. Third, the refined knowledge is used to retrain the cross-modal aligner, yielding more robust medical image representations while mitigating bias and noise introduced by other modalities. Comprehensive evaluations on brain, bladder, and liver cancer datasets demonstrate that KAMP outperforms existing methods in most downstream few-shot classification tasks.

Reward models (RMs) are essential for training large language models (LLMs), but remain underexplored for omni models that handle interleaved image and text sequences. We introduce Multimodal RewardBench 2 (MMRB2), the first comprehensive benchmark for reward models on multimodal understanding and (interleaved) generation. MMRB2 spans four tasks: text-to-image, image editing, interleaved generation, and multimodal reasoning ("thinking-with-images"), providing 1,000 expert-annotated preference pairs per task from 23 models and agents across 21 source tasks. MMRB2 is designed with: (1) practical but challenging prompts; (2) responses from state-of-the-art models, and agents; and (3) preference pairs with strong human-expert consensus, curated via an ensemble filtering strategy. Using MMRB2, we study existing judges for each subtask, including multimodal LLM-as-a-judge and models trained with human preferences. Top judges like GPT-5 and Gemini-2.5-Pro reach 66-75% accuracy, compared to >90% for humans, and outperform the commonly used GPT-4o (59% accuracy). The best performing open-source model Qwen3-VL-32B achieves similar accuracies as Gemini-2.5-Flash (64%). We also show that MMRB2 performance strongly correlates with downstream task success, and conduct an in-depth analysis that shows key areas to improve the reward models going forward.

High-precision dichotomous image segmentation (DIS) is a task of extracting fine-grained objects from high-resolution images.Existing methods trade efficiency for accuracy: non-diffusion methods are fast but suffer from weak semantics and unstable spatial priors, causing false detections; diffusion-based methods offer high accuracy via strong generative priors but are computationally expensive.In depth maps, a complete object appears as a low variance region with a smooth interior and sharp boundaries, whereas the background exhibits a chaotic, high variance pattern due to disconnected surfaces at varying depths. We refer to this as the depth integrity-prior.Inspired by this, and noting that DIS currently lacks depth maps, we leverage pseudo-depth information from monocular depth estimation models to obtain essential semantic understanding, thereby rapidly revealing spatial differences across target objects and the background.To exploit this prior, we propose the Prior-guided Depth Fusion Network (PDFNet), which fuses RGB and pseudo-depth features for depth-aware structure perception. We further introduce a novel depth integrity-prior loss to enforce depth consistency in segmentation and a fine-grained enhancement module with adaptive patch selection to sharpen boundaries.Notably, PDFNet with DAM-v2 achieves SOTA (F^ max _b 0.915 on DIS-VD and 0.915 on DIS-TE) using less than half the params of diffusion-based methods.