CVPR 2026 论文列表

Embodied Question Answering (EQA) requires agents to navigate 3D environments, accumulate visual evidence, and reason over partial observations to answer questions. However, current agents struggle to maintain coherent, long-horizon behavior: planning remains reactive, causing inconsistent actions, while monolithic memories entangle all observations, hindering retrieval of the sparse but crucial evidence. We address these issues by reframing EQA through the lens of predictive processing, in which coherent behavior emerges from a prediction-correction loop grounded in stable priors. Guided by this perspective, we propose Predict Before You Explore (Pred-EQA), an architecture that integrates predictive planning with specialized memory. A high-level planner predicts where question-relevant evidence is likely to appear and generates a compact set of actionable exploration branches encoding long-horizon intent. A low-level executor then reduces uncertainty within these branches, revising predictions when they fail. A dual-memory system complements this process by separating slowly evolving structural priors from compact, question-relevant visual evidence, enabling consistent planning and efficient evidence accumulation. Through this prediction-guided exploration, Pred-EQA achieves coherent trajectories under partial observability. Experiments on OpenEQA and Express-Bench show that Pred-EQA achieves state-of-the-art results in both accuracy and exploration efficiency, demonstrating the benefits of prediction-driven embodied reasoning.

In stroke-based rendering, search methods often get trapped in local minima due to discrete stroke placement, while differentiable optimizers lack structural awareness and produce unstructured layouts. To bridge this gap, we propose a dual representation that couples discrete polylines with continuous Bezier control points via a bidirectional mapping mechanism. This enables collaborative optimization: local gradients refine global stroke structures, while content-aware stroke proposals help escape poor local optima. Our representation further supports Gaussian-splatting-inspired initialization, enabling highly parallel stroke optimization across the image. Experiments show that our approach reduces the number of strokes by 30-50%, achieves more structurally coherent layouts, and improves reconstruction quality, while cutting optimization time by 30-40% compared to existing differentiable vectorization methods.

Tokenization is a fundamental technique in the generative modeling of various modalities. In particular, it plays a critical role in autoregressive (AR) models, which have recently emerged as a compelling option for 3D generation.However, optimal tokenization of 3D shapes remains an open question. State-of-the-art (SOTA) methods primarily rely on geometric level-of-detail (LoD) hierarchies, originally designed for rendering and compression. These spatial hierarchies are often token-inefficient and lack semantic coherence for AR modeling. We propose Level-of-Semantics Tokenization (LoST), which orders tokens by semantic salience, such that early prefixes decode into complete, plausible shapes that possess principal semantics, while subsequent tokens refine instance-specific geometric and semantic details. To train LoST, we introduce Relational Inter-Distance Alignment (RIDA), a novel 3D semantic alignment loss, inspired by relational knowledge distillation, that aligns the relational structure of the 3D shape latent space with that of the semantic DINO feature space. Experiments show that LoST achieves SOTA reconstruction, surpassing previous LoD-based 3D shape tokenizers by large margins on both geometric and semantic reconstruction metrics. Moreover, LoST achieves efficient, high-quality AR 3D generation and enables downstream tasks like semantic retrieval, while using only 0.1%-10% of the tokens needed by prior 3D AR models. Code is available via the project website at https://lost3d.github.io/

We investigate state-of-the-art deepfake detectors that leverage ViT-based vision foundation models and discover that the [CLS] token suffers from the Pre-trained Information Bias (PIB), i.e., it tends to mainly focus on global semantics due to the knowledge dominated by pre-trained model parameters, while struggling to emphasize subtle local forgery cues. To overcome this limitation, one potential way is incorporating the token-level features to reform a detection-specific token. To this end, we propose Query-Driven Token-Level Forgery Purification (QTFP) framework to better capture local forgery traces without losing useful pre-trained prior. Specifically, we introduce randomly initialized, learnable query tokens independent of the backbone and prior knowledge, which effectively aggregate multi-patch evidence into a global token for detection. To make query tokens focus on meaningful regions, we propose a theoretical fake-likelihood contrastive learning loss, which employs a weighting strategy to highlight significant fake regions while diminishing real-like patch impact. Using SNR theory, we verify that the designed weight is both reliable and informative. To further maintain useful authentic information, a real-attention alignment constraint is applied to query tokens. These designs go beyond relying solely on the [CLS] token by jointly reorganizing real and fake information across all tokens, which successfully enhance detector robustness. Extensive experiments on diverse datasets demonstrate the effectiveness of our method.

We suggest a new multi-modal algorithm for joint inference of paired structurally aligned samples with Rectified Flow models. While some existing methods propose a codependent generation process, they do not view the problem of joint generation from a structural alignment perspective. Recent work uses Score Distillation Sampling to generate aligned 3D models, but SDS is known to be time-consuming, prone to mode collapse, and often provides cartoonish results. By contrast, our suggested approach relies on the joint transport of a segment in the sample space, yielding faster computation at inference time. Our approach can be built on top of an arbitrary Rectified Flow model operating on the structured latent space. We show the applicability of our method to the domains of image, video, and 3D shape generation using state-of-the-art baselines and evaluate it against both editing-based and joint inference-based competing approaches. We demonstrate a high degree of structural alignment for the sample pairs obtained with our method and a high visual quality of the samples. Our method improves the state-of-the-art for image and video generation pipelines. For 3D generation, it is able to show comparable quality while working orders of magnitude faster.

Capturing dynamic spatiotemporal neural activity is essential for understanding large-scale brain mechanisms. Functional magnetic resonance imaging (fMRI) provides high-resolution cortical representations that form a strong basis for characterizing fine-grained brain activity patterns. The high acquisition cost of fMRI limits large-scale applications, therefore making high-quality fMRI reconstruction a crucial task. Electroencephalography (EEG) offers millisecond-level temporal cues that complement fMRI. Leveraging this complementarity, we present an EEG-conditioned framework for reconstructing dynamic fMRI as continuous neural sequences with high spatial fidelity and strong temporal coherence at the cortical-vertex level. To address sampling irregularities common in real fMRI acquisitions, we incorporate a null-space intermediate-frame reconstruction, enabling measurement-consistent completion of arbitrary intermediate frames and improving sequence continuity and practical applicability. Experiments on the CineBrain dataset demonstrate superior voxel-wise reconstruction quality and robust temporal consistency across whole-brain and functionally specific regions. The reconstructed fMRI also preserves essential functional information, supporting downstream visual decoding tasks. This work provides a new pathway for estimating high-resolution fMRI dynamics from EEG and advances multimodal neuroimaging toward more dynamic brain activity modeling.

Large vision-language models struggle with medical video understanding, where spatial precision, temporal reasoning, and clinical semantics are critical. To address this, we first introduce MedVidBench, a large-scale benchmark of 531,850 video-instruction pairs across 8 medical sources spanning video, segment, and frame-level tasks, curated through a rigorous quality assurance pipeline with expert-guided prompting and dual-model validation. While supervised fine-tuning on MedVidBench yields noticeable gains, standard Reinforcement Learning (RL) fails due to imbalanced reward scales across datasets, which destabilizes optimization and leads to training collapse. To overcome this, we introduce MedGRPO, a novel RL framework for balanced multi-dataset training with two key innovations: (1) cross-dataset reward normalization that maps each dataset's median performance to a common reward value, ensuring fair optimization regardless of difficulty, and (2) a medical LLM judge that evaluates caption quality on five clinical dimensions through comparative similarity scoring. Supervised fine-tuning Qwen2.5-VL-7B on MedVidBench substantially outperforms GPT-4.1 and Gemini-2.5-Flash across all tasks, demonstrating MedVidBench's efficacy, while our MedGRPO framework further improves upon the SFT baseline across grounding and captioning tasks. Our work establishes a foundational benchmark and robust training methodology for advancing vision-language models in medical domains.

Accurate camera viewpoint estimation under sparse-view conditions remains challenging, particularly in two-view scenarios. Recent approaches leverage diffusion models such as Zero123 to synthesize novel views conditioned on relative viewpoint, showing promising results when repurposed for viewpoint estimation via optimization with MSE loss. However, existing methods often suffer from non-convex loss landscape with numerous local minima, making them sensitive to initialization and reliant on naive multi-start strategies. We analyze these optimization challenges and visualize failure cases, showing that geometric ambiguities, such as symmetry and self-similarity, can mislead gradient-based updates toward incorrect viewpoints. To address these limitations, we propose a score-based method that reshapes the optimization landscape to guide updates toward the ground-truth viewpoint, followed by a refinement stage using a viewpoint-conditioned diffusion model. Experiments show that our method improves convergence, reduces reliance on brute-force sampling, and achieves competitive accuracy with higher sample-efficiency.

We introduce OLATverse, a large-scale dataset comprising around 9M images of 765 real-world objects, captured from multiple viewpoints under a diverse set of precisely controlled lighting conditions. While recent advances in object-centric inverse rendering, novel view synthesis and relighting have shown promising results, most techniques still heavily rely on the synthetic datasets for training and small-scale real-world datasets for benchmarking, which limits their realism and generalization. To address this gap, OLATverse offers two key advantages over existing datasets: large-scale coverage of real objects and high-fidelity appearance under precisely controlled illuminations. Specifically, OLATverse contains 765 common and uncommon real-world objects, spanning a wide range of material categories. Each object is captured using 35 DSLR cameras and 331 individually controlled light sources, enabling the simulation of diverse illumination conditions. In addition, for each object, we provide well-calibrated camera parameters, accurate object masks, photometric surface normals, and diffuse albedo as auxiliary resources. We also construct an extensive evaluation set, establishing the first comprehensive real-world object-centric benchmark for inverse rendering and normal estimation. OLATverse is publicly available at https://vcai.mpi-inf.mpg.de/projects/OLATverse/.

Visual autoregressive models achieve remarkable generation quality through next-scale predictions across multi-scale token pyramids. However, the conventional method uses uniform scale downsampling to build these pyramids, leading to aliasing artifacts that compromise fine details and introduce unwanted jaggies and moire patterns. To tackle this issue, we present FVAR, which reframes the paradigm from next-scale prediction to next-focus prediction, mimicking the natural process of camera focusing from blur to clarity. Our approach introduces three key innovations: 1) Next-Focus Prediction Paradigm that transforms multi-scale autoregression by progressively reducing blur rather than simply downsampling; 2) Progressive Refocusing Pyramid Construction that uses physics-consistent defocus kernels to build clean, alias-free multi-scale representations; and 3) High-Frequency Residual Learning that employs a specialized residual teacher network to effectively incorporate alias information during training while maintaining deployment simplicity. Specifically, we construct optical low-pass views using defocus point spread function (PSF) kernels with decreasing radius, creating smooth blur-to-clarity transitions that eliminate aliasing at its source. To further enhance detail generation, we introduce a High-Frequency Residual Teacher that learns from both clean structure and alias residuals, distilling this knowledge to a vanilla VAR deployment network for seamless inference. Extensive experiments on ImageNet demonstrate that FVAR substantially reduces aliasing artifacts, improves fine detail preservation, and enhances text readability, achieving superior performance with perfect compatibility to existing VAR frameworks.

Instruction-based multimodal image manipulation has recently made rapid progress. However, existing evaluation methods lack a systematic and human-aligned framework for assessing model performance on complex and creative editing tasks. To address this gap, we propose CREval, a fully automated question-answer (QA)-based evaluation pipeline that overcomes the incompleteness and poor interpretability of opaque Multimodal Large Language Models (MLLMs) scoring. Simultaneously, we introduce CREval-Bench, a comprehensive benchmark specifically designed for creative image manipulation under complex instructions. CREval-Bench covers three categories and nine creative dimensions, comprising over 800 editing samples and 13K evaluation queries. Leveraging this pipeline and benchmark, we systematically evaluate a diverse set of state-of-the-art open and closed-source models. The results reveal that while closed-source models generally outperform open-source ones on complex and creative tasks, all models still struggle to complete such edits effectively. In addition, user studies demonstrate strong consistency between CREval's automated metrics and human judgments. Therefore, CREval provides a reliable foundation for evaluating image editing models on complex and creative image manipulation tasks, and highlights key challenges and opportunities for future research.

Tracking any point (TAP) is a fundamental yet challenging task in computer vision, requiring high precision and long-term motion reasoning. Recent attempts to combine RGB frames and event streams have shown promise, yet they typically rely on synchronous or non-adaptive fusion, leading to temporal misalignment and severe degradation when one modality fails. We introduce TAPFormer, a transformer-based framework that performs asynchronous temporal-consistent fusion of frames and events for robust and high-frequency arbitrary point tracking.Our key innovation is a Transient Asynchronous Fusion (TAF) mechanism, which explicitly models the temporal evolution between discrete frames through continuous event updates--bridging the gap between low-rate frames and high-rate events. In addition, a Cross-modal Locally Weighted Fusion (CLWF) module adaptively adjusts spatial attention according to modality reliability, yielding stable and discriminative features even under blur or low light.To evaluate our approach under realistic conditions, we construct a novel real-world frame-event TAP dataset under diverse illumination and motion conditions.Our method outperforms existing point trackers, achieving a 28.2% improvement in average pixel error within threshold. Moreover, on standard point tracking benchmarks, our tracker consistently achieves the best performance. We will release the code and dataset upon acceptance to support future research.

Beyond general recognition tasks, specialized domains and fine-grained settings often encounter data scarcity, especially for tail classes. To obtain less biased and more reliable models under such scarcity, practitioners leverage diffusion models to supplement underrepresented regions of real data. Specifically, recent studies fine-tune pretrained diffusion models with LoRA on few-shot real sets to synthesize additional images. While an image-wise LoRA trained on a single image captures fine-grained details yet offers limited diversity, a class-wise LoRA trained over all shots produces diverse images as it encodes class priors yet tends to overlook fine details. To combine both benefits, we separate the adapter into a class-shared LoRA A for class priors and per-image LoRAs \mathcal B for image-specific characteristics. To expose coherent class semantics in the shared LoRA A, we propose a semantic boosting by preserving class bounding boxes during training. For generation, we compose A with a mixture of \mathcal B using coefficients drawn from a Dirichlet distribution. Across diverse datasets, our synthesized images are both diverse and detail-rich while closely aligning with the few-shot real distribution, yielding robust gains in downstream classification accuracy.

Hyperbolic spaces provide a natural geometry for representing hierarchical and tree-structured data due to their exponential volume growth. To leverage these benefits, neural networks require intrinsic and efficient components that operate directly in hyperbolic space. In this work, we lift two core components of neural networks, Multinomial Logistic Regression (MLR) and Fully Connected (FC) layers, into hyperbolic space via Busemann functions, resulting in Busemann MLR (BMLR) and Busemann FC (BFC) layers with a unified mathematical interpretation. BMLR provides compact parameters, a point-to-horosphere distance interpretation, batch-efficient computation, and a Euclidean limit, while BFC generalizes FC and activation layers with comparable complexity. Experiments on image classification, genome sequence learning, node classification, and link prediction demonstrate improvements in effectiveness and efficiency over prior hyperbolic layers. The code is available at https://github.com/GitZH-Chen/HBNN.

Model merging aims to combine multiple task-specific experts into a single model, but inter-task interference often causes severe degradation, especially when the experts are trained on heterogeneous objectives. Existing data-free methods are practical, yet largely rely on parameter-space heuristics without explicitly modeling task statistics. In this paper, we show that under a local linear approximation, the input covariance of each task can be estimated directly from its fine-tuning update, providing a principled bridge between parameter changes and data geometry in the data-free setting. Based on this insight, we propose ACE-Merging, a closed-form model merging framework with adaptive covariance normalization, a collective structural prior, and spectral refinement. Across both vision and language benchmarks, ACE-Merging achieves strong overall performance among data-free baselines. In particular, it improves the average score on GPT-2 by more than 4 points over prior methods, while also delivering strong scalability and competitive efficiency. Our code is available at https://github.com/unravel-xu/ACE-Merging/tree/main.

The deployment of multimodal models in high-stakes domains, such as self-driving vehicles and medical diagnostics, demands not only strong predictive performance but also reliable mechanisms for detecting failures. In this work, we address the largely unexplored problem of failure detection in multimodal contexts. We propose Adaptive Confidence Regularization (ACR), a novel framework specifically designed to detect multimodal failures. Our approach is driven by a key observation: in most failure cases, the confidence of the multimodal prediction is significantly lower than that of at least one unimodal branch, a phenomenon we term confidence degradation. To mitigate this, we introduce an Adaptive Confidence Loss that penalizes such degradations during training. In addition, we propose Multimodal Feature Swapping, a novel outlier synthesis technique that generates challenging, failure-aware training examples. By training with these synthetic failures, ACR learns to more effectively recognize and reject uncertain predictions, thereby improving overall reliability. Extensive experiments across four datasets, three modalities, and multiple evaluation settings demonstrate that ACR achieves consistent and robust gains. The source code will be available at https://github.com/mona4399/ACR.

Recent diffusion-based one-step methods have shown remarkable progress in the field of image super-resolution, yet they remain constrained by three critical limitations: (1) inferior fidelity performance caused by the information loss from compression encoding of low-quality (LQ) inputs; (2) insufficient region-discriminative activation of generative priors; (3) misalignment between text prompts and their corresponding semantic regions. To address these limitations, we propose CODSR, a controllable one-step diffusion network for image super-resolution. First, we propose an LQ-guided feature modulation module that leverages original uncompressed information from LQ inputs to provide high-fidelity conditioning for the diffusion process. We then develop a region-adaptive generative prior activation method to effectively enhance perceptual richness without sacrificing local structural fidelity. Finally, we employ a text-matching guidance strategy to fully harness the conditioning potential of text prompts. Extensive experiments demonstrate that CODSR achieves superior perceptual quality and competitive fidelity compared with state-of-the-art methods while maintaining efficient one-step inference.

Pre-trained text-to-image (T2I) generative priors have shown success in depth and normal prediction. However, dense prediction is inherently an image-to-image task, suggesting that image editing models, rather than T2I generative models, may be a more suitable foundation for fine-tuning. Motivated by this, we conduct a systematic analysis of the fine-tuning behaviors of both editors and generators for dense geometry estimation. Our findings show that editing models possess inherent structural priors, which enable them to converge more stably by "refining" their innate features, and ultimately achieve higher performance than their generative counterparts. Based on these findings, we introduce FE2E, a framework that pioneeringly adapts an advanced editing model based on Diffusion Transformer (DiT) architecture for dense geometry prediction. Specifically, to tailor the editor for this deterministic task, we reformulate the editor's original flow matching loss into the "consistent velocity" training objective. And we use logarithmic quantization to resolve the precision conflict between the editor's native BFloat16 format and the high precision demand of our tasks. Additionally, we repurpose the editor's discarded region for a cost-free joint estimation of depth and normals, which improves the inference efficiency. Without scaling up the training data, FE2E achieves impressive performance improvements in zero-shot monocular depth and normal estimation across multiple datasets. Notably, it achieves over 35% performance gains on the ETH3D dataset and outperforms the DepthAnything series, which is trained on 100xdata.

The demand for dynamic 3D assets in AR/VR has recently popularized Deformable Gaussian Splatting. However, traditional RGB cameras are limited in their ability to reconstruct high-speed scenes due to motion blur and low temporal resolution. While event cameras offer a promising alternative, reconstructing a complete scene from their sparse and noisy output is a significant challenge. Existing event-based methods rely on an auxiliary sensor, such as a frame camera, thereby inducing tedious hardware and calibration challenges.We introduce FastEventDGS, a novel Deformable Gaussian Splatting-based framework that leverages a single event camera for high-fidelity 4D reconstruction. Our method utilizes a continuous camera trajectory parametrization and integrates two event generation models to provide both photometric and geometric constraints. We further propose a local patch event motion loss to constrain object motion, effectively mitigating overfitting. To ensure robust reconstruction, we employ an off-the-shelf model for depth correction and apply noise regularization terms in the final stage. We demonstrate robust results on both new synthetic and real-world datasets, highlighting our framework's ability to provide a simplified, event-only solution for high-fidelity 4D reconstruction in dynamic scenes.