CVPR 2026 论文列表

Diffusion-based text-to-image (T2I) models have enabled remarkable generative capabilities, yet precise text-based image editing that preserves the original's structural and perceptual fidelity remains non-trivial. Existing approaches either rely on retraining with large bespoke datasets, incurring significant computational and curation costs, or adopt lightweight fine-tuning strategies that still require optimization and often fail in fine-grained or semantically complex edits. We propose NEAF (Natural image Editing with Attention Fusion), a novel zero-shot, universal tuning-free framework for arbitrary T2I models, obviating the need for dataset curation or retraining. NEAF introduces a lightweight, learnable XA-Conductor module that dynamically identifies salient cross-attention contributions pertinent to the edit. This module optimizes a weight vector to orchestrate an adaptive fusion of cross-attention maps derived from the source, edited, and reconstruction branches. This triadic-feedback optimization strategy ensures the precise instantiation of user directives while rigorously preserving the fidelity of quiescent regions. Extensive experiments validate NEAF as a flexible and general framework that consistently surpasses existing methods across diverse editing tasks, demonstrating particular dominance in complex, non-rigid editing scenarios where other approaches falter.

Continual learning for video--language understanding is increasingly important as models face non-stationary data, domains, and query styles, yet prevailing solutions blur what should stay stable versus what should adapt, rely on static routing/capacity, or require replaying past videos. We aim to explicitly specify where stability lives and where plasticity should be focused under realistic memory and privacy constraints. We introduce Affordance-First Decomposition (AFD): videos are mapped to slowly varying affordance tokens that form a shared, time-aligned substrate, while a lightweight, query-routed, conflict-aware scheduler concentrates adaptation and grows capacity only when needed. The substrate is stabilized via weak alignment and teacher consistency, and training uses question-only replay. AFD achieves state-of-the-art across protocols: 51.6% average accuracy with -1.8% forgetting on domain-incremental VideoQA, ViLCo R@1@0.5 of 29.6% (MQ) and 20.7% (NLQ) with 18.4% stAP@0.25 (VQ), and 39.5% accuracy with -1.6% forgetting on time-incremental iVQA. Overall, AFD offers an explicit, interpretable split between a stable interaction-centered substrate and targeted adaptation.

Instruction-based image editing through natural language has emerged as a powerful paradigm for intuitive visual manipulation. While recent models achieve impressive results on single edits, they suffer from severe quality degradation under multi-turn editing. Through systematic analysis, we identify progressive loss of high-frequency information as the primary cause of this quality degradation. We present FreqEdit, a training-free framework that enables stable editing across 10+ consecutive iterations. Our approach comprises three synergistic components: (1) high-frequency feature injection from reference velocity fields to preserve fine-grained details, (2) an adaptive injection strategy that spatially modulates injection strength for precise region-specific control, and (3) a path compensation mechanism that periodically recalibrates the editing trajectory to prevent over-constraint. Extensive experiments demonstrate that FreqEdit achieves superior performance in both identity preservation and instruction following compared to seven state-of-the-art baselines.

While recent advancements in generative models have achieved remarkable visual fidelity in video synthesis, creating coherent multi-shot narratives remains a significant challenge. To address this, keyframe-based approaches have emerged as a promising alternative to computationally intensive end-to-end methods, offering the advantages of fine-grained control and greater efficiency. However, these methods often fail to maintain cross-shot consistency and capture cinematic language. In this paper, we introduce STAGE, a SToryboard-Anchored GEneration workflow to reformulate the keyframe-based multi-shot video generation task. Instead of using sparse keyframes, we propose STEP^2 to predict a structural storyboard composed of start-end frame pairs for each shot. We introduce the multi-shot memory pack to ensure long-range entity consistency, the dual-encoding strategy for intra-shot coherence, and the two-stage training scheme to learn cinematic inter-shot transition. We also contribute the large-scale ConStoryBoard dataset, including high-quality movie clips with fine-grained annotations for story progression, cinematic attributes, and human preferences. Extensive experiments demonstrate that STAGE achieves superior performance in structured narrative control and cross-shot coherence. Our code will be available at this url.

Many recent human motion prediction methods adopt a multi-stage refinement framework, where each stage produces an initial guess of future poses for the next stage. These guesses are progressively refined towards the target prediction through a sequence of spatial-temporal reasoning stages.However, such a cascaded design incurs large computation and memory overheads that grow at least linearly with network depth, and lack an explicit stopping criteria.In this paper, we propose MotionDEQ, a deep equilibrium motion predictor that reformulates progressive guessing paradigm as a fixed point problem within an implicit layer. This formulation is conceptually equivalent to performing infinitely many refinement steps, but requires only O(1) training memory and can be solved efficiently through any black-box solvers. We carefully design this implicit refinement process by integrating Euclidean geometric transformations into equilibrium learning, allowing the entire network to be equivariant. We also find DEQs naturally fit the real-world scenario where motion data comes streamingly: the converged fixed point can be reused as a warm initial guess, to help recycle the redundant inference computation when making subsequent predictions.Our experiments demonstrate that MotionDEQ achieves the state-of-the-art prediction performances with superior memory efficiency, using fewer than 300K parameters with 55.3mm prediction error at 400ms on the Human3.6M dataset.

Distributed Image Compression (DIC) is crucial for multi-view transmission, especially when operating at extremely low bitrates (< 0.1 bpp). Its core challenge is effectively utilizing side information to achieve high-quality reconstruction under strict bitrate budgets. However, existing DIC approaches struggle to exploit global context and object-level details from side information, leading to local blurring and the loss of fine details in the reconstruction. To address these limitations, we propose a Multimodal DIC framework (MDIC), which, for the first time, leverages side information in a multimodal manner into the DIC paradigm, effectively preserving fine-grained local details and enhancing global perceptual quality in reconstructed images. Specifically, we introduce a text-to-image diffusion-based decoder conditioned on textual side information extracted from correlated images to capture shared global semantics. Moreover, we design a feature-mask generator, supervised by a multimodal fine-grained alignment task, to strengthen the exploitation of visual side information. The generated mask serves two purposes: first, it guides the extraction of fine-grained details from losslessly transmitted side information to preserve the semantic consistency of reconstructed details; second, it regulates the extraction of clustered feature representations from the quantized VQ-VAE embeddings, compensating for category information lost under the extreme compression of the primary image. Extensive experiments on the widely used KITTI Stereo and Cityscapes datasets demonstrate that MDIC achieves state-of-the-art perceptual quality at extremely low bitrates.

Recent progress in GPU-accelerated, photorealistic simulation has opened a scalable data-generation path for robot learning, where massive physics and visual randomization allow policies to generalize beyond curated environments. Building on these advances, we develop a teacher-student-bootstrap learning framework for vision-based humanoid loco-manipulation, using articulated-object interaction as a representative high-difficulty benchmark. Our approach introduces a staged-reset exploration strategy that stabilizes long-horizon privileged-policy training, and a GRPO-based fine-tuning procedure designed to mitigate partial observability and improve closed-loop consistency in sim-to-real RL. Trained entirely on synthetic simulation data, the resulting policy achieves robust zero-shot performance across diverse articulated objects--including multiple door types--and outperforms human teleoperators by up to 31.7% in task completion time under the same whole-body control stack. This represents the first humanoid sim-to-real policy capable of diverse articulated loco-manipulation from pure RGB perception.

The neural process (NP) is a probabilistic meta learning model that learns distributions over functions via a global latent variable.It enables fast adaptation in few-shot scenarios by leveraging past experience. However, the design of latent variable structures and conditioning mechanisms in NPs remains underexplored, despite their importance in capturing diverse functional distributions.This paper proposes a new variant of NPs via mixture density modeling, referred to as the neural mixture density process (NMDP).The NMDP decomposes model parameters into task-agnostic and task-specific components to represent function distributions more flexibly. We train the model via the Expectation-Maximization algorithm to construct expressive functional priors.Compared with existing work, our method maintains several advantages: (i) less overfitting by updating a small part of the network parameters, (ii) compact task representation via distributions in the simplex,(iii) an improvement guarantee of generative likelihoods over iteration. Experimental results show that our method can achieve competitive performance with adequate explainability.

Vision-Language-Action (VLA) models trained via imitation learning suffer from significant performance degradation in data-scarce scenarios due to their reliance on large-scale demonstration datasets. Although reinforcement learning (RL)-based post-training has proven effective in addressing data scarcity, its application to VLA models is hindered by the non-resettable nature of real-world environments. This limitation is particularly critical in high-risk domains such as industrial automation, where interactions often induce state changes that are costly or infeasible to revert. Furthermore, existing VLA approaches lack a reliable mechanism for detecting task completion, leading to redundant actions that reduce overall task success rates. To address these challenges, we propose RehearseVLA, an RL-based post-training framework that replaces physical interaction with a low-cost world model-based virtual simulator. RehearseVLA consists of two key components: (1) a physically-consistent world simulator that generates temporally consistent future visual observations, and (2) a vision-language model (VLM)-guided instant reflector that provides continuous reward signals and predicts action termination. This simulated environment enables VLA models to safely explore and generalize beyond their initial imitation learning distribution. Our method achieves notable performance gains with as few as five expert demonstrations per task. Experiments on complex robotic manipulation tasks demonstrate that RehearseVLA effectively overcomes the data inefficiency, safety constraints, and inefficient execution of conventional VLA models that rely on real-world interaction, offering a practical and scalable solution for post-training in resource-constrained settings. Our code is available at https://github.com/iSEE-Laboratory/RehearseVLA.

Although current watermarking techniques for 3D Gaussian Splatting (3DGS) are promising in protecting the copyrights of both 3DGS models and their rendered images, they greatly suffer from low watermark robustness and poor rendering quality when applying quantization to large 3DGS models to accommodate resource-limited devices. To address these problems, this paper introduces a novel two-stage quantization-aware 3DGS watermarking approach called Robust3DGSW. By properly embedding watermarks into the mid-frequency bands of both the 3D Gaussian parameters and 2D rendered images, the first stage of Robust3DGSW can effectively counteract the quantization-induced signal loss and mitigate the adverse effects of watermarks on rendered images. In the second stage, Robust3DGSW trains both 2D and 3D decoders using our proposed multi-scale adversarial perturbation approach, alongside a gradual quantization process, which enables robust watermark extraction even under excessive quantization. Comprehensive experimental results obtained from the well-known Blender, LLFF, and MipNeRF-360 datasets demonstrate that, when compared to leading 3DGS watermarking techniques, Robust3DGSW not only mitigates the negative effects of quantization on watermarks but also enables fast rendering with high quality.

Modern industrial quality control heavily relies on automated anomaly detection. While few-shot anomaly detection addresses the challenge of limited labeled data, real-world inspection faces a vast diversity of anomaly types, sizes, and shapes. We identify the primary cause for the anomaly detection difficulty as the progressive loss of detect cues as they pass through deep feature extraction pipelines. To counteract the defect cue fading, we propose a Defect Cue-Preserved Structural Feature Refinement model, referred to as DCP-SFR. Recognizing that early-stage cues are paramount, we design a conditional anomaly cue amplification module to produce an initial anomaly score map, which is then enhanced to increase the contrast between anomalous and normal regions. The amplified cues is subsequently used for reconstruction-based anomaly localization, by anchoring attention on true anomaly regions to preserve spatial integrity and prevent drift. Further, we incorporate a structure-aware segmentation refinement stage to improve anomaly segmentation in terms of edge alignment, thereby significantly improve boundary accuracy. On the MVTec AD and VisA benchmarks, DCP-SFR achieves state-of-the-art performance, with an image-level AUROC of 97.3% and a pixel-level AUROC of 98.2%, demonstrating strong cross-domain generalization performance.

Diffusion models have achieved remarkable performance on a wide range of generative tasks, yet training them from scratch is notoriously resource-intensive, typically requiring millions of training images and many GPU days. Motivated by a data-centric view of this bottleneck, we adopt a condensation-based perspective: given a large training set, the goal is to construct a much smaller condensed dataset that still supports training strong diffusion models under minimal data and compute budgets. To operationalize this perspective, we introduce Diffusion Dataset Condensation (D2C), a two-phase framework comprising Select and Attach. In the Select phase, a diffusion difficulty score combined with interval sampling is used to identify a compact, informative training subset from the original data. Building on this subset, the Attach phase further strengthens the conditional signals by augmenting each selected image with rich semantic and visual representations. To our knowledge, D2C is the first framework that systematically investigates dataset condensation for diffusion models, whereas prior condensation methods have mainly targeted discriminative architectures. Extensive experiments across data budgets from 0.8% to 8% of ImageNet, model architectures, and image resolutions demonstrate that D2C dramatically accelerates diffusion model training while preserving high generative quality. On ImageNet 256x256 with SiT-XL/2, D2C attains a FID of 4.3 in just 40k steps using only 0.8% of the training images, corresponding to about 233x and 100x faster training than vanilla SiT-XL/2 and SiT-XL/2 + REPA, respectively. Code: https://github.com/xie-lab-ml/Diffusion-Dataset-Condensation.

Image-level Weakly Supervised Semantic Segmentation (WSSS) typically leverages Class Activation Maps (CAMs) for pixel-wise localization. However, existing CLIP-based methods often yield under-activated CAMs, primarily due to the inaccurate semantic relationships in the affinity-based refinement. In this work, we propose a novel framework, CD-CLIP (Class Distribution based CLIP), which addresses this issue by introducing a Class Distribution Aware (CDA) module. The CDA module captures richer semantic relationships by modeling patch-wise distributions across all classes using Jensen-Shannon divergence, thereby enhancing the completeness of CAMs. While this significantly improves the coverage of the foreground class, the over-activation at class boundaries might also exist due to the comprehensive integration of relationships between inter target classes. To mitigate this adverse effect on segmentation supervision, we introduce a Super-class Boundary Exploration (SBE) module, which leverages structural knowledge of DINO to generate boundary-aware super-class prototype CAMs. By employing the boundary-enhanced loss, our SBE module effectively provides accurate boundary supervision for the final segmentation. Our proposed CD-CLIP framework achieves state-of-the-art performance on both PASCAL VOC and MS COCO benchmarks. Code will be released.

Dataset distillation often prioritizes global semantic proximity when creating small surrogate datasets for original large-scale ones. However, object semantics are inherently hierarchical. For example, the position and appearance of a bird's eyes are constrained by the outline of its head. Global proximity alone fails to capture how object-relevant structures at different levels support recognition. In this work, we investigate the contributions of hierarchical semantics to effective distilled data. We leverage the vision autoregressive (VAR) model whose coarse-to-fine generation mirrors this hierarchy and propose HierAmp to amplify semantics at different levels. At each VAR scale, we inject class tokens that dynamically identify salient regions and use their induced maps to guide amplification at that scale. This adds only marginal inference cost while steering synthesis toward discriminative parts and structures. Empirically, we find that semantic amplification leads to more diverse token choices in constructing coarse-scale object layouts. Conversely, at fine scales, the amplification concentrates token usage, increasing focus on object-related details. Across popular dataset distillation benchmarks, HierAmp consistently improves validation performance without explicitly optimizing global proximity, demonstrating the importance of semantic amplification for effective dataset distillation.

Given a monocular video, the goal of video re-rendering is to generate views of the scene from a novel camera trajectory. Existing methods face two distinct challenges. Geometrically unconditioned models lack spatial awareness, leading to drift and deformation under viewpoint changes. On the other hand, geometrically-conditioned models depend on estimated depth and explicit reconstruction, making them susceptible to depth inaccuracies and calibration errors.We propose to address these challenges by using the implicit geometric knowledge embedded in the latent space of a large 4D reconstruction model to condition the video generation process. These latents capture scene structure in a continuous space without explicit reconstruction. Therefore, they provide a flexible representation that allows the pretrained diffusion prior to regularize errors more effectively. By jointly conditioning on these latents and source camera poses, we demonstrate that our model achieves state-of-the-art results on the video re-rendering task.

High-quality data has become a primary driver of progress under scale laws, with curated datasets often outperforming much larger unfiltered ones at lower cost. Online data curation extends this idea by dynamically selecting training samples based on the model's evolving state. While effective in classification and multimodal learning, existing online sampling strategies rarely extend to object detection because of its structural complexity and domain gaps. We introduce DetGain, an online data curation method specifically for object detection that estimates the marginal perturbation of each image to dataset-level Average Precision (AP) based on its prediction quality. By modeling global score distributions, DetGain efficiently estimates the global AP change and computes teacher-student contribution gaps to select informative samples at each iteration. The method is architecture-agnostic and minimally intrusive, enabling straightforward integration into diverse object detection architectures. Experiments on the COCO dataset with multiple representative detectors show consistent improvements in accuracy. DetGain also demonstrates strong robustness under low-quality data and can be effectively combined with knowledge distillation techniques to further enhance performance, highlighting its potential as a general and complementary strategy for data-efficient object detection.

Theory of Mind (ToM) refers to the ability to infer others' mental states, such as beliefs, desires, and intentions. Current vision-language embodied agents lack ToM-based decision-making, and existing benchmarks focus solely on human mental states while ignoring the agent's own perspective, hindering coherent decision and action generation. To address this, we propose MindPower, a Robot-Centric framework integrating Perception, Mental Reasoning, Decision Making and Action. Given multimodal inputs, MindPower first perceives the environment and human states, then performs ToM Reasoning to model both self and others, and finally generates decisions and actions guided by inferred mental states. Furthermore, we introduce Mind-Reward, a novel optimization objective that encourages VLMs to produce consistent ToM Reasoning and behavior. Our model outperforms GPT-4o by 12.77% in decision making and 12.49% in action generation. Project page: https://zhangdaxia22.github.io/MindPower/.

Large multimodal models (LMMs) exhibit strong task generalization capabilities, offering new opportunities for zero-shot visual anomaly segmentation (ZSAS). However, existing LMM-based segmentation approaches still face fundamental limitations: anomaly concepts are inherently abstract and context-dependent, lacking stable visual prototypes, and the weak alignment between high-level semantic embeddings and pixel-level spatial features hinders precise anomaly localization. To address these challenges, we present AG-VAS (Anchor-Guided Visual Anomaly Segmentation), a new framework that expands the LMM vocabulary with three learnable semantic anchor tokens--[SEG], [NOR], and [ANO], establishing a unified anchor-guided segmentation paradigm. Specifically, [SEG] serves as an absolute semantic anchor that translates abstract anomaly semantics into explicit, spatially grounded visual entities (e.g., holes or scratches), while [NOR] and [ANO] act as relative anchors that model the contextual contrast between normal and abnormal patterns across categories. To further enhance cross-modal alignment, we introduce a Semantic-Pixel Alignment Module (SPAM) that aligns language-level semantic embeddings with high-resolution visual features, along with an Anchor-Guided Mask Decoder (AGMD) that performs anchor-conditioned mask prediction for precise anomaly localization. In addition, we curate Anomaly-Instruct20K, a large-scale instruction dataset that organizes anomaly knowledge into structured descriptions of appearance, shape, and spatial attributes, facilitating effective learning and integration of the proposed semantic anchors. Extensive experiments on six industrial and medical benchmarks demonstrate that AG-VAS achieves consistent state-of-the-art performance in the zero-shot setting.

Stochastic Gradient Descent (SGD) and its momentum variants form the backbone of deep learning optimization, yet the underlying dynamics of their gradient behavior remain insufficiently understood. In this work, we reinterpret gradient updates through the lens of signal processing and reveal that fixed momentum coefficients inherently distort the balance between bias and variance, leading to skewed or suboptimal parameter updates. To address this, we propose SGDF (SGD with Filter), an optimizer inspired by the principles of Optimal Linear Filtering. SGDF computes an online, time-varying gain to dynamically refine gradient estimation by minimizing the mean-squared error, thereby achieving an optimal trade-off between noise suppression and signal preservation. Furthermore, our approach could extend to other optimizers, showcasing its broad applicability to optimization frameworks. Extensive experiments across diverse architectures and benchmarks demonstrate SGDF surpasses conventional momentum methods and achieves performance on par with or surpassing state-of-the-art optimizers.