CVPR 2026 Papers

Recent breakthroughs in 3D generative modeling have yielded remarkable progress in static shape synthesis, yet truly dynamic 4D generation remains elusive, hindered by temporal artifacts and prohibitive computational demand. We present Sculpt4D, a native 4D generative framework that seamlessly integrates efficient temporal modeling into a pretrained 3D Diffusion Transformer (Hunyuan3D 2.1), thereby mitigating the scarcity of 4D training data. At its core lies a Block Sparse Attention mechanism that preserves object identity by anchoring to the initial frame while capturing rich motion dynamics via a time-decaying sparse mask. This design faithfully models complex spatiotemporal dependencies with high fidelity, while sidestepping the quadratic overhead of full attention and reducing network total computation by 56%. Consequently, Sculpt4D establishes a new state-of-the-art in temporally coherent 4D synthesis, charting a path toward efficient and scalable 4D generation.

Dynamic multimodal fusion methods lack robust theoretical guidance for handling modal conflicts and inconsistent data quality. While recent theoretical work has associated modal weights with scalar quantities such as loss or confidence, this paradigm struggles to comprehensively address the risks caused by inconsistent predictive distributions across modalities. In this paper, we propose a **Co**nflict-driven **Ri**sk **M**inimization (**CoRiM**) dynamic fusion paradigm. Specifically, we redefine dynamic fusion as a principled, per-sample, direct risk minimization task. To this end, we first design a novel, differentiable Modality Conflict Risk (MCR) function, \mathcal R (w), which quantifies risk by directly modeling fused uncertainty and inter-modal consistency. Second, we identify that minimizing \mathcal R (w) is fundamentally a non-convex constrained optimization problem over the probabilistic simplex. To efficiently solve this specific challenge, we innovatively introduce the projection free Frank-Wolfe (FW) algorithm, as it is perfectly suited for optimization on the simplex.We prove that our designed \mathcal R (w) possesses L-smoothness, which provides theoretical guarantees for the convergence of the FW algorithm on our non-convex objective. Extensive experiments on multiple benchmark datasets demonstrate that CoRiM outperforms current state-of-the-art methods in high-conflict and noisy environments, validating the robustness of our method.

Leveraging 3D information within Multimodal Large Language Models (MLLMs) has recently shown significant advantages for indoor scene understanding. However, existing methods, including those using explicit ground-truth 3D positional encoding and those grafting external 3D foundation models for implicit geometry, struggle with the trade-off in 2D-3D representation fusion, leading to suboptimal deployment. To this end, we propose 3D-Implicit Depth Emergence, a method that reframes 3D perception as an emergent property derived from geometric self-supervision rather than explicit encoding. Our core insight is the Implicit Geometric Emergence Principle: by strategically leveraging privileged geometric supervision through mechanisms like a fine-grained geometry validator and global representation constraints, we construct an information bottleneck. This bottleneck forces the model to maximize the mutual information between visual features and 3D structures, allowing 3D awareness to emerge naturally within a unified visual representation. Unlike existing approaches, our method enables 3D perception to emerge implicitly, disentangling features in dense regions and, crucially, eliminating depth and pose dependencies during inference with zero latency overhead. This paradigm shift from external grafting to implicit emergence represents a fundamental rethinking of 3D knowledge integration in visual-language models. Extensive experiments demonstrate that our method surpasses SOTA on multiple 3D scene understanding benchmarks. Our approach achieves a 55% reduction in inference latency while maintaining strong performance across diverse downstream tasks, underscoring the effectiveness of meticulously designed auxiliary objectives for dependency-free 3D understanding. Source code can be found at \href https://chushanzhang.github.io/3D-IDE/ \textcolor[HTML] E8477A \small\texttt github.com/ChushanZhang/3D-IDE .

Hierarchical Vision-Language-Action (VLA) models have rapidly become a dominant paradigm for robotic manipulation. It typically comprising a Vision-Language backbone for perception and understanding, together with a generative policy for action generation. However, its performance is increasingly bottlenecked by the action generation proceess. (i) Low inference efficiency. A pronounced distributional gap between isotropic noise priors and target action distributions, which increases denoising steps and the incidence of infeasible samples. (ii) Poor robustness. Existing policies condition solely on the current observation, neglecting the constraint of history sequence and thus lacking awareness of task progress and temporal consistency. To address these issues, we introduce OptimusVLA, a dual-memory VLA framework with Global Prior Memory (GPM) and Local Consistency Memory (LCM). GPM replaces Gaussian noise with task-level priors retrieved from semantically similar trajectories, thereby shortening the generative path and reducing the umber of function evaluations (NFE). LCM dynamically models executed action sequence to infer task progress and injects a learned consistency constraint that enforces temporal coherence and smoothness of trajectory. Across three simulation benchmarks, OptimusVLA consistently outperforms strong baselines: it achieves 98.6% average success rate on LIBERO, improves over pi_0 by 13.5% on CALVIN, and attains 38% average success rate on RoboTwin 2.0 Hard. In Real-World evaluation, OptimusVLA ranks best on Generalization and Long-horizon suites, surpassing pi_0 by 42.9% and 52.4%, respectively, while delivering 2.9x inference speedup.

Recently, masked skeleton reconstruction models have emerged as strong action representation learners, driving significant progress in self-supervised skeleton-based action recognition. However, existing state-of-the-art methods must predict an exceedingly large number of spatiotemporal patches, significantly prolonging training time. Besides, by treating all spatiotemporal regions equally during reconstruction, these models are distracted from learning the critical motion patterns that underlie action semantics. To address these challenges, we propose Adaptive Masked Reconstruction (AMR), a faster and stronger pre-training framework. We first decouple the decoder from the encoder, enabling flexible prediction of larger spatiotemporal patches and dramatically reducing reconstruction complexity. Given that larger patches contain more complex information, which is challenging to predict and consequently degrades performance, we accordingly introduce an adaptive guidance module. This module identifies regions of high motion informativeness, guiding the model to focus on the most discriminative parts of each patch and alleviating reconstruction difficulty. Experiments on NTU RGB+D 60, NTU RGB+D 120, and PKU-MMD datasets demonstrate that AMR not only accelerates pre-training substantially but also improves downstream recognition accuracy, surpassing current state-of-the-art approaches.

Single-image 3D human reconstruction holds significant promise due to its convenience and high demand in various applications. Previous methods have garnered tremendous progress by employing 2D multi-view diffusion models to generate auxiliary views as reconstruction priors, but they struggle with 3D inconsistencies and limited generalization capabilities. In this paper, we present FISHuman, which aims to generate fine-grained, high-fidelity, and content-wise diverse 3D humans from a single-view input, providing production-ready 3D assets. We propose an elaborately designed workflow that reconstructs dynamic 3D meshes from multi-view inconsistent guidance. Specifically, we adapt a dual-stream transformer-based video diffusion model to generate cross-modally aligned multi-view RGB and normal sequences. We find that naively employing static 3D reconstruction can lead to geometric distortions and texture blurriness, due to the lack of 3D awareness within the generated frames. To address this, we introduce a novel 4D remeshing module that explicitly disentangles the learning of the globally shared canonical mesh and transient variations by tracking per-vertex deformations under different viewpoints. The topological consistency of the deformed meshes inherently enables the optimization of a unified UV representation that effectively integrates appearance attributes across frames. Both qualitative and quantitative experimental results demonstrate the superiority of our method over prior works in terms of appearance realism, geometric fineness, and generalization diversity. We also showcase the applicability of our reconstructed avatars for downstream applications including animation and 3D editing.

We present BioCoach, a biomechanics-grounded vision-language framework for fitness coaching from streaming video. BioCoach fuses two signals, visual appearance and 3D skeletal kinematics, through a novel three-stage pipeline: an exercise-specific degree-of-freedom selector that focuses analysis on salient joints; a structured biomechanical context that pairs individualized morphometrics with cycle and constraint analysis; and a vision-biomechanics conditioned feedback module that applies cross-attention to generate precise, actionable text. Using parameter-efficient training that freezes the vision and language backbones, BioCoach yields transparent, personalized reasoning rather than pattern matching. To enable learning and fair evaluation, we augment QEVD-fit-coach with biomechanics-oriented feedback to create QEVD-bio-fit-coach, and we introduce a biomechanics-aware LLM judge metric. BioCoach delivers clear gains on QEVD-bio-fit-coach across lexical and judgment metrics while maintaining temporal triggering; on the original QEVD-fit-coach, it improves text quality and correctness with near-parity timing, demonstrating that explicit kinematics and constraints are key to accurate, phase-aware coaching.

Point cloud completion is an important yet challenging problem in 3D computer vision, which aims to reconstruct complete and dense 3D shapes from partial point clouds. Although transformer-based and geometry-based approaches have made significant progress, they often struggle to capture the complex, high-order correlations inherent in point clouds. To address this limitation, we propose Hyper-PCN, a point cloud completion framework that leverages hypergraphs to explicitly model complex, higher-order correlations within incomplete inputs for more accurate completion. It comprises two key modules: Hyper Refinement Stack, designed to progressively capture coarse-to-fine high-order correlations through a series of hypergraph learning stages, and Anchor-based Hypergraph Neural Network, which employs a two-stage sampling strategy to construct collaborative hypergraphs, ensuring robust modeling of global structures. Extensive experiments on multiple datasets demonstrate that our approach consistently outperforms state-of-the-art methods.

Smart glass is emerging as an useful device since it provides plenty of insights under hands-busy, eyes-on-task situations. To understand the context of the wearer, 6D object pose estimation in egocentric view is becoming essential. However, existing 6D object pose estimation benchmarks fail to capture the challenges of real-world egocentric applications, which are often dominated by severe motion blur, dynamic illumination, and visual obstructions. This discrepancy creates a significant gap between controlled lab data and chaotic real-world application. To bridge this gap, we introduce EgoXtreme, a new large-scale 6D pose estimation dataset captured entirely from an egocentric perspective. EgoXtreme features three challenging scenarios--industrial maintenance, sports, and emergency rescue--designed to introduce severe perceptual ambiguities through extreme lighting, heavy motion blur, and smoke. Evaluations of state-of-the-art generalizable pose estimators on EgoXtreme indicate that their generalization fails to hold in extreme conditions, especially under low light. We further demonstrate that simply applying image restoration (e.g., deblurring) offers no positive improvement for extreme conditions. While performance gain has appeared in tracking-based approach, implying using temporal information in fast-motion scenarios is meaningful. We conclude that EgoXtreme is an essential resource for developing and evaluating the next generation of pose estimation models robust enough for real-world egocentric vision. The dataset and code are available at https://taegyoun88.github.io/EgoXtreme/

Federated Learning (FL) serves as a prominent distributed training paradigm, enabling devices to collaboratively train a shared model with local data. However, in practice, clients generally possess data from multiple distributional domains, posing significant challenges to efficiency and generalization. In this paper, we propose a domain-sensitive federated pruning framework, FedFIP, that preserves domain-invariant structures while retaining domain-specific representations. The Domain-Sensitive Fisher Pruning (DSFP) module estimates channel importance per domain via Fisher information, and uploads this signal to the server to obtain a globally shared pruning mask. Given local domain heterogeneity, each client reuses its Fisher information to selectively reactivate domain-specific channels, yielding personalized sparse models that remain structurally aligned yet adapt to local heterogeneity. To further enhance performance, we adopt a Domain-Sensitive Regularization (DSR) module: the server builds domain prototypes from uploaded importance signals and broadcasts them back. Guided by the domain prototypes, we adopt a structure-contrastive loss to strengthen intra-domain consistency and inter-domain discriminability. We propose a structure-aware aggregation algorithm that fuses heterogeneous personalized architectures into a domain-generalized global model. Experiments on multi-domain benchmarks demonstrate that FedFIP surpasses state-of-the-art FL baselines while substantially shrinking model size.

We introduce a test-time framework for multiview Transformers (MVTs) that incorporates priors (e.g., camera poses, intrinsics, and depth) to improve 3D tasks without retraining or modifying pre-trained image-only networks. Rather than feeding priors into the architecture, we cast them as constraints on the predictions and optimize the network at inference time. The optimization loss consists of a self-supervised objective and prior penalty terms. The self-supervised objective captures the compatibility among multi-view predictions and is implemented using photometric or geometric loss between renderings from other views and each view itself. Any available priors are converted into penalty terms on the corresponding output modalities. Across a series of 3D vision benchmarks, including point map estimation and camera pose estimation, our method consistently improves performance over base MVTs by a large margin. On the ETH3D, 7-Scenes, and NRGBD datasets, our method reduces the point-map distance error by more than half compared with the base image-only models. Our method also outperforms retrained prior-aware feed-forward methods, demonstrating the effectiveness of our test-time constrained optimization (TCO) framework for incorporating priors into 3D vision tasks. The code is available at https://github.com/cvlab-stonybrook/TCO.

Modern vision-language pipelines are driven by RGB vision encoders trained on massive image-text corpora. While these pipelines have enabled impressive zero-shot capabilities and strong transfer across tasks, they still inherit two structural inefficiencies from the pixel domain: (i) transmitting dense RGB images from edge devices to the cloud is energy-intensive and costly, and (ii) patch-based tokenization explodes sequence length, stressing attention budgets and context limits. We explore 2D Gaussian Splatting (2DGS) as an alternative visual substrate for alignment: a compact, spatially adaptive representation that parameterizes images by a set of colored anisotropic Gaussians. We develop a scalable 2DGS pipeline with structured initialization, luminance-aware pruning, and batched CUDA kernels, achieving over 90x faster fitting and 97% GPU utilization compared to prior implementations. We further adapt contrastive language-image pre-training (CLIP) to 2DGS by reusing a frozen RGB-based transformer backbone with a lightweight splat-aware input stem and a perceiver resampler, training only 9.7 - 13.8% of the total parameters. On a 12.8M dataset from DataComp, GS encoders yield competitive zero-shot performance on 38 datasets from the CLIP benchmark while compressing inputs 3-23.5x relative to pixels. When integrated into a full end-to-end LLaVA-style visual-language model (VLM), GS encoders consistently outperform the RGB VLM baseline across six VQA benchmarks, often with reduced token counts, demonstrating that a transmission-efficient, structured Gaussian representation can match or surpass pixel-based models in downstream multimodal reasoning. Our codebase can be found at https://github.com/ Tambe-Lab/GaussianVision.

Face recognition remains vulnerable to presentation attacks, calling for robust Face Anti-Spoofing (FAS) solutions. Recent MLLM-based FAS methods reformulate the binary classification task as the generation of brief textual descriptions to improve cross-domain generalization. However, their generalizability is still limited, as such descriptions mainly capture intuitive semantic cues (e.g., mask contours) while struggling to perceive fine-grained visual patterns. To address this limitation, we incorporate external visual tools into MLLMs to encourage deeper investigation of subtle spoof clues. Specifically, we propose the Tool-Augmented Reasoning FAS (TAR-FAS) framework, which reformulates the FAS task as a Chain-of-Thought with Visual Tools (CoT-VT) paradigm, allowing MLLMs to begin with intuitive observations and adaptively invoke external visual tools for fine-grained investigation. To this end, we design a tool-augmented data annotation pipeline and construct the ToolFAS-16K dataset, which contains multi-turn tool-use reasoning trajectories. Furthermore, we introduce a tool-aware FAS training pipeline, where Diverse-Tool Group Relative Policy Optimization (DT-GRPO) enables the model to autonomously learn efficient tool use. Extensive experiments under a challenging one-to-eleven cross-domain protocol demonstrate that TAR-FAS achieves SOTA performance while providing fine-grained visual investigation for trustworthy spoof detection.

Recent advances in diffusion models have enabled high-fidelity, identity-preserving image generation for personalized applications such as digital avatars and virtual try-on systems. However, their reliance on sensitive facial reference images raises growing privacy concerns. Existing defense mechanisms are primarily designed for training-based personalization and struggle to generalize to emerging training-free approaches, due to fundamental differences in their identity integration paradigms. To bridge this gap, we propose IDGuardian--the first generalizable and model-agnostic identity protection framework capable of defending against both training-based and training-free personalization methods. IDGuardian abstracts the personalization process into two critical stages: identity extraction and identity injection. It then introduces crafted adversarial perturbations to simultaneously disrupt both stages. Specifically, it degrades the identity features extracted by external encoders and establishes an adversarial conceptual bridge that misdirects the generative trajectory away from the target identity. Extensive experiments show that IDGuardian effectively protects identity across various personalization pipelines and model architectures, while remaining robust to post-processing, adaptive attacks, and cross-dataset generalization.

Large-scale image datasets frequently contain identifiable or sensitive content, raising privacy risks when training models that may memorize and leak such information. We present Unsafe2Safe, a fully automated pipeline that detects privacy-prone images and rewrites only their sensitive regions using multimodally guided diffusion editing. Unsafe2Safe operates in two stages. Stage 1 uses a vision-language model to (i) inspect images for privacy risks, (ii) generate paired private and public captions that respectively include and omit sensitive attributes, and (iii) prompt a large language model to produce structured, identity-neutral edit instructions conditioned on the public caption. Stage 2 employs instruction-driven diffusion editors to apply these dual textual prompts, producing privacy-safe images that preserve global structure and task-relevant semantics while neutralizing private content. To measure anonymization quality, we introduce a unified evaluation suite covering Quality, Cheating, Privacy, and Utility dimensions. Across MS-COCO, Caltech101 and MIT Indoor67, Unsafe2Safe reduces face similarity, text similarity, and demographic predictability by large margins, while maintaining downstream model accuracy comparable to training on raw data. Fine-tuning diffusion editors on our automatically generated triplets (private caption, public caption, edit instruction) further improves both privacy protection and semantic fidelity. Unsafe2Safe provides a scalable, principled solution for constructing large, privacy-safe datasets without sacrificing visual consistency or downstream utility.

Space-time video super-resolution (STVSR) is a task aimed at simultaneously upsampling a video in both spatial and temporal dimensions. Previous studies on STVSR have primarily focused on task-specific architectures and modeling paradigms, while effective pretraining strategies remain underexplored. In this paper, we propose a pseudo-temporal space-time reconstruction pretraining framework for STVSR networks that enables effective use of image datasets, which naturally provide strong spatial cues. Each training sample is constructed by duplicating a single image into a pseudo-temporal video and independently zero-filling random pixel regions across its frames. Instead of designing a separate pretraining module, we pretrain the STVSR network on a task aligned with its core objectives of spatial restoration and cross-frame aggregation. The model learns to reconstruct clean, higher-spatio-temporal-resolution outputs from degraded, pseudo-temporal inputs, with a modulation factor encouraging greater focus on difficult regions. Extensive experiments show that our simple pretraining significantly improves STVSR performance and outperforms existing video representation learning approaches. We note our method is effective even when pretraining and finetuning with a limited quantity of data.

Multimodal large language models (MLLMs) have achieved remarkable success across diverse applications, from autonomous driving to document understanding. As these models are deployed in safety-critical contexts, understanding their adversarial robustness becomes crucial. However, current evaluations focus primarily on simple tasks like coarse-grained classification, and employ inconsistent evaluation protocols, hindering rigorous comparison of attack methods. We introduce AdvRobustBench, a comprehensive adversarial robustness benchmark for MLLMs comprising 1,000 examples across visual question answering (VQA) and optical character recognition (OCR) tasks, drawn from widely-used MLLM benchmarks (MMBench, MMStar, OCRBench-v2). We further propose Omni-Attack, a novel transfer-based black-box attack method that addresses key challenges in attacking open-ended question-answering systems. Our approach introduces (i) a target-construction pipeline that generates question-conditioned textual and visual targets to provide stronger optimization signals, and (ii) a location-aware attack strategy for OCR that enables spatially-precise perturbations. Extensive experiments demonstrate that Omni-Attack achieves strong targeted attack success rates (up to 71.8% on GPT-4.1 at \varepsilon=8/255) across both proprietary models (GPT-4.1, Claude 3.7, Gemini 2.0) and open-source MLLMs, revealing significant vulnerabilities in current multimodal systems. Our benchmark and findings establish a foundation for developing more robust MLLMs. Codes are available at https://github.com/hukkai/transferable_mllm_attack

Talking head generation creates lifelike avatars from static portraits for virtual communication and content creation. However, current models do not yet convey the feeling of truly interactive communication, often generating one-way responses that lack emotional engagement. We identify two key challenges toward truly interactive avatars: generating motion in real-time under causal constraints and learning expressive, vibrant reactions without additional labeled data. To address these challenges, we propose Avatar Forcing, a new framework for interactive head avatar generation that models real-time user-avatar interactions through diffusion forcing. This design allows the avatar to process real-time multimodal inputs, including the user's audio and motion, with low latency for instant reactions to both verbal and non-verbal cues such as speech, nods, and laughter. Furthermore, we introduce a direct preference optimization method that leverages synthetic losing samples constructed by dropping user conditions, enabling label-free learning of expressive interaction. Experimental results demonstrate that our framework enables real-time interaction with low latency (about 500ms), achieving 6.8x speedup compared to the baseline, and produces reactive and expressive avatar motion, which is preferred over 80% against the baseline.

We introduce GeoSAM2, a prompt-controllable framework for 3D part segmentation that casts the task as multi-view 2D mask prediction. Given a textureless object, we render normal and point maps from predefined viewpoints and accept simple 2D prompts--clicks or boxes--to guide part selection. These prompts are processed by a shared SAM2 backbone augmented with LoRA and residual geometry fusion, enabling view-specific reasoning while preserving pretrained priors. The predicted masks are back-projected to the object, aggregated across views.Our method enables fine-grained, part-specific control without requiring text prompts, per-shape optimization, or full 3D labels. In contrast to global clustering or scale-based methods, prompts are explicit, spatially grounded, and interpretable. We achieve state-of-the-art class-agnostic performance on PartObjaverse-Tiny and PartNetE, outperforming both slow optimization-based pipelines and fast but coarse feedforward approaches. Our results highlight a new paradigm: aligning the paradigm of 3D segmentation with SAM2, leveraging interactive 2D inputs to unlock controllability and precision in object-level part understanding.