CVPR 2026 Papers

Traditional classifiers treat all class labels as mutually independent, thereby considering all negative classes to be equally incorrect. This approach fails severely in many real-world scenarios, where a known semantic hierarchy defines a partial order of preferences over negative classes. While hierarchy-aware feature representations have shown promise in mitigating this problem, their performance is typically assessed using metrics like Mistake Severity (MS) and Average Hierarchical Distance (AHD). In this paper, we highlight important shortcomings in existing hierarchical evaluation metrics, demonstrating that they are often incapable of measuring true hierarchical performance. Our analysis reveals that existing methods learn sub-optimal hierarchical representations, despite competitive MS and AHD scores. To counter these issues, we introduce Hierarchical Composition of Orthogonal Subspaces (Hier-COS), a novel framework for unified 'hierarchy-aware fine-grained' and 'hierarchical multi-label' classification. We show that Hier-COS is theoretically guaranteed to be consistent with the given hierarchy tree. Furthermore, our framework implicitly adapts the learning capacity for different classes based on their position within the hierarchy tree -- a vital property absent in existing methods. Finally, to address the limitations of evaluation metrics, we propose Hierarchically Ordered Preference Score (HOPS), a ranking-based metric that demonstrably overcomes the deficiencies of current evaluation standards. We benchmark Hier-COS on four challenging datasets, including the deep and imbalanced tieredImageNet-H (12-level) and iNaturalist-19 (7-level). Through extensive experiments, we demonstrate that Hier-COS achieves state-of-the-art performance across all hierarchical metrics for every dataset, while simultaneously beating the top-1 accuracy in all but one case. Lastly, we show that Hier-COS can effectively learn to transform the frozen features extracted from a pretrained backbone (ViT) to be hierarchy-aware, yielding substantial benefits for hierarchical classification performance.

Recent advances in multimodal large language models (MLLMs) have accelerated progress in domain-oriented AI, yet their development in geoscience and remote sensing (RS) remains constrained by distinctive challenges: wide-ranging disciplinary knowledge, heterogeneous sensor modalities, and a fragmented spectrum of tasks. To bridge these gaps, we introduce GeoMMBench, a comprehensive multimodal question-answering benchmark covering diverse RS disciplines, sensors, and tasks, enabling broader and more rigorous evaluation than prior benchmarks. Using GeoMMBench, we assess 36 open-source and proprietary large language models (LLMs), uncovering systematic deficiencies in domain knowledge, perceptual grounding, and reasoning--capabilities essential for expert-level geospatial interpretation. Beyond evaluation, we propose GeoMMAgent, a multi-agent framework that strategically integrates retrieval, perception, and reasoning through domain-specific RS models and tools. Extensive experimental results demonstrate that GeoMMAgent significantly outperforms standalone LLMs, underscoring the importance of tool-augmented agents for dynamically tackling complex geoscience and RS challenges.

Advertising images significantly impact commercial conversion rates and brand equity, yet current evaluation methods rely on subjective judgments, lacking scalability, standardized criteria, and interpretability. To address these challenges, we present **A3 (Advertising Aesthetic Assessment)**, a comprehensive framework encompassing four components: a paradigm (**A3-Law**), a dataset (**A3-Dataset**), a multimodal large language model (**A3-Align**), and a benchmark (**A3-Bench**). Central to A3 is a theory-driven paradigm, A3-Law, comprising three hierarchical stages: (1) Perceptual Attention, evaluating perceptual image signals for their ability to attract attention; (2) Formal Interest, assessing formal composition of image color and spatial layout in evoking interest; and (3) Desire Impact, measuring desire evocation from images and their persuasive impact. Building on A3-Law, we construct A3-Dataset with 120K instruction-response pairs from 30K advertising images, each richly annotated with multi-dimensional labels and Chain-of-Thought (CoT) rationales. We further develop A3-Align, trained under A3-Law with CoT-guided learning on A3-Dataset. Extensive experiments on A3-Bench demonstrate that A3-Align achieves superior alignment with A3-Law compared to existing models, and this alignment generalizes well to quality advertisement selection and prescriptive advertisement critique, indicating its potential for broader deployment. Dataset, code, and models can be found at: https://github.com/euleryuan/A3-Align

Gaze estimation is instrumental in modern virtual reality (VR) systems. Despite significant progress in remote-camera gaze estimation, VR gaze research remains constrained by data scarcity, particularly the lack of large-scale, accurately labeled datasets captured with the off-axis camera configurations typical of modern headsets. Gaze annotation is difficult since fixation on intended targets cannot be guaranteed. To address these challenges, we introduce VRGaze, the first large-scale off-axis gaze estimation dataset for VR, comprising 2.1 million near-eye infrared images collected from 68 participants. We further propose GazeShift, an attention-guided unsupervised framework for learning gaze representations without labeled data. Unlike prior redirection-based methods that rely on multi-view or 3D geometry, GazeShift is tailored to near-eye imagery, achieving effective gaze-appearance disentanglement in a compact, real-time model. GazeShift embeddings can be optionally adapted to individual users via lightweight few-shot calibration, achieving a 1.84deg mean error on VRGaze. On the remote-camera MPIIGaze dataset, the model achieves a 7.15deg person-agnostic error, doing so with 10x fewer parameters and 35x fewer FLOPs than baseline methods. Deployed natively on a VR headset GPU, inference takes only 5 ms. Combined with demonstrated robustness to illumination changes, these results highlight GazeShift as a label-efficient, real-time solution for VR gaze tracking. Project code and the VRGaze dataset are released at https://github.com/gazeshift3/gazeshift

We introduce a new template-free tracking paradigm based solely on natural language, capable of tracking an arbitrary object and seamlessly switching to a new target without box initialization.Our key idea is to localize an object via vision-language (VL) correlation.However, using the correlation alone is brittle under large search regions due to spatial uncertainty and ambiguous VL saliency. To resolve these, we propose MVLM, a memory-based vision-language margin confidence that integrates vision-language correlation, encoder prediction, and temporal memory.MVLM dynamically gates the search region--switching between compact ROI (Region of Interest) search and global re-localization--to reduce spatial uncertainty. Theoretically, we derive bounds that connect the MVLM score to tracking probability, characterizing mis-localization within ROI and ROI-exclusion probabilities.Through extensive evaluation, we validate our theorems and achieve state-of-the-art performance on several benchmarks (TNL2K, LaSOT, OTB99 and MGIT) using only language guidance.

Inspired by the success of Reinforcement Learning with Human Feedback (RLHF) in image generation, recent work has adapted reward-based learning to image super-resolution (ISR) by using Image Quality Assessment (IQA) models as rewards. However, existing IQA models typically output only a single global score and are insensitive to local, fine-grained distortions, allowing perceptually undesirable artifacts to obtain spuriously high rewards and leading to reward hacking. To address this issue, we propose FinPercep-RM, a fine-grained perceptual reward model built on an encoder-decoder architecture that predicts both a global quality score and a Perceptual Degradation Map for spatially localizing and quantifying local defects. We further introduce FGR-30k, a dataset containing diverse and subtle distortions produced by real-world super-resolution models, to train the reward model. While FinPercep-RM provides stronger supervision, its increased complexity also makes generator policy learning unstable. We therefore develop a Co-evolutionary Curriculum Learning (CCL) strategy, in which the reward model and the ISR model evolve synchronously: the reward signal progressively increases in complexity, while the ISR model starts with simple global supervision for fast convergence and gradually transitions to fine-grained rewards. This easy-to-hard design stabilizes training and suppresses reward hacking. Extensive experiments across multiple ISR models demonstrate improvements in both global quality and local realism. Code will be available at https://github.com/lyd-2022/FinPercep-RM.

Model merging offers a promising paradigm for consolidating multiple expert models into a single multitask architecture. However, its effectiveness is often hindered by task interference, where conflicting parameter updates from different tasks degrade performance. While dynamic, Mixture-of-Experts based methods have improved adaptability, they are fundamentally limited by constructing their expert pools from task vectors in isolation, failing to resolve underlying structural conflicts across tasks. In this paper, we introduce DuetMerging, a novel framework that synergistically mitigates task interference from both dynamic and static perspectives. Dynamically, we apply Tucker decomposition to a unified tensor of task vectors, creating a harmonized expert pool derived from a shared core tensor that structurally enhances synergies and suppresses conflicts. Statically, we introduce a neuron-based sparsification technique that leverages task-specific neuron activation patterns to create a precise mask. This allows us to selectively preserve critical information from the decomposition's residual while suppressing functionally irrelevant or conflicting parameters. Comprehensive experiments demonstrate that DuetMerging outperforms existing methods, establishing a new state-of-the-art in both task performance and parameter efficiency.

Extreme low-light Raw image restoration remains challenging due to overwhelming noise and severe detail loss.In this paper, we exploit the potential of the dual-exposure setting for this severely ill-posed problem.Existing methods suffer from unreliable cross-exposure alignment, resulting in degraded detail recovery and compromised color fidelity. To address these challenges, we propose RawMetaDiff, a novel generative diffusion framework that restores a high-fidelity Raw image from a short-exposure input, conditioned on a potentially misaligned long-exposure reference under the guidance of Raw metadata.At its core, we propsed two complementary mechanisms: the Meta-Assistant Color Transfer (MACT) enforces color consistency by aligning global color statistics along the channel dimension,while the Meta-Normed Cross Attention (MNCA) leverages Raw metadata to establish robust cross-exposure spatial correspondences and inject shadow details.To support robust diffusion training, we first collect a 1K real-world, dual-exposure Raw dataset, namely DERaw, and then design a realistic degradation model to synthesize data that closely approximates real-world conditions.Extensive experiments on both synthetic and real-world datasets demonstrate that RawMetaDiff significantly outperforms existing methods, justifying an effective new solution for extreme low-light Raw image restoration from the generative perspective.

To enhance the perception and reasoning capabilities of multimodal large language models in complex visual scenes, recent research has introduced agent-based workflows. In these works, MLLMs autonomously utilize image cropping tool to analyze regions of interest for question answering. While existing training strategies, such as those employing supervised fine-tuning and reinforcement learning, have made significant progress, our empirical analysis reveals a key limitation. We demonstrate the model's strong reliance on global input and its weak dependence on the details within the cropped region. To address this issue, we propose a novel two-stage reinforcement learning framework that does not require trajectory supervision. In the first stage, we introduce the "Information Gap" mechanism by adjusting the granularity of the global image. This mechanism trains the model to answer questions by focusing on cropped key regions, driven by the information gain these regions provide. The second stage further enhances cropping precision by incorporating a grounding loss, using a small number of bounding box annotations. Experiments show that our method significantly enhances the model's attention to cropped regions, enabling it to achieve state-of-the-art performance on high-resolution visual question-answering benchmarks. Our method provides a more efficient approach for perceiving and reasoning fine-grained details in MLLMs. Code is available at: https://github.com/XuanPu-Z/LFPC.

Instruction tuning has been central to the success of recent vision-language models (VLMs), but it remains expensive-requiring large-scale datasets, high-quality annotations, and large compute budgets. We propose PRioritized cOncept learninG via Relative Error-driven Sample Selection (PROGRESS), a data- and compute-efficient framework that enables VLMs to dynamically select what to learn next based on their evolving needs during training. At each stage, the model tracks its learning progress across skills and selects the most informative samples-those it has not already mastered and that are not too difficult to learn at the current stage of training. This strategy effectively controls skill acquisition and the order in which skills are learned. Specifically, we sample from skills showing the highest learning progress, prioritizing those with the most rapid improvement. Unlike prior methods, PROGRESS requires no upfront answer annotations, queries answers only on a need basis, avoids reliance on additional supervision from auxiliary VLMs, and does not require compute-heavy gradient computations for data selection. Experiments across multiple instruction-tuning datasets of varying scales demonstrate that PROGRESS consistently outperforms state-of-the-art baselines with much less data and supervision. Additionally, we show strong cross-architecture generalization and transferability to larger models, validating PROGRESS as a scalable solution for efficient learning.

Monocular Semantic Scene Completion (SSC) aims to reconstruct complete 3D semantic scenes from a single RGB image, offering a cost-effective solution for autonomous driving and robotics. However, the inherently imbalanced nature of voxel distributions--where over 93% of voxels are empty and foreground classes are rare--poses significant challenges. Existing methods often suffer from redundant emphasis on uninformative voxels and poor generalization to long-tailed categories. To address these issues, we propose VoxSAMNet (Voxel Sparsity-Aware Modulation Network), a unified framework that explicitly models voxel sparsity and semantic imbalance. Our approach introduces: (1) a Dummy Shortcut for Feature Refinement (DSFR) module that bypasses empty voxels via a shared dummy node while refining occupied ones with deformable attention; (2) a Foreground Modulation Strategy combining Foreground Dropout (FD) and Text-Guided Image Filter (TGIF) to alleviate overfitting and enhance class-relevant features. Extensive experiments on the public benchmarks SemanticKITTI and SSCBench-KITTI-360 demonstrate that VoxSAMNet achieves state-of-the-art performance, surpassing prior monocular and stereo baselines with mIoU scores of 18.2% and 20.2%, respectively. Our results highlight the importance of sparsity-aware and semantics-guided design for efficient and accurate 3D scene completion, offering a promising direction for future research.

Tracking-any-point (TAP) answers query-conditioned correspondence but leaves the dense, all-pairs structure of a video implicit. We formulate All-Pairs Tracking (APT): given a video, predict dense displacement and visibility for every source-target frame pair, from which per-pixel trajectories can be read out. To this end, we propose PairFormer, a feed-forward transformer that addresses APT in a single pass. A spatio-temporal patch encoder computes temporally conditioned features for all frames. \revaddtwo CorrBank builds a learnable correlation bank for each frame pair and produces pairwise motion tokens. A broadcast motion mixer aggregates trajectory-wise context and broadcasts it back to refine the pairwise motion tokens. A trajectory head first predicts coarse dense displacement, visibility, and confidence, and then refines them iteratively to form a coherent all-pairs trajectory field. To support APT at scale, we develop PAIRender, a data platform that synthesizes photo-realistic dynamic scenes with dense annotations. From PAIRender we derive a training set (\pi-R10K) and a benchmark (APT-Bench) with an all-to-all evaluation protocol. Experiments show that PairFormer achieves strong performance on APT-Bench and competitive results on standard TAP benchmarks. Code and dataset will be released upon publication.

Depth priors provide salient geometric structure that benefits camouflaged object detection (COD), but directly using Monocular Depth Estimation (MDE) causes a task misalignment that still fails to identify camouflaged objects.To address this issue, we propose the Depth Segment Anything Model (DepthSAM), a MDE-adapted method specifically designed to mitigate this misalignment.DepthSAM incorporates two core innovations: (1) a Sparse Mixture-of-Experts Adapter (SMEA) that enables MDE to learn semantic information unique to camouflaged scenes, and (2) a Geometric-Semantic Fusion Module (GSFM) that efficiently integrates geometric cues with high-level semantics. With these components, DepthSAM achieves both robust semantic understanding in camouflaged environments and accurate segmentation of camouflaged objects.Extensive experiments show that DepthSAM achieves new SOTA performance on three major benchmarks. For example, on COD10K, its S_ \alpha and F_ b ^ \omega metrics surpass the best competing methods by 3.0% and 4.3%, respectively.

Recent subject-driven image customization excels in fidelity, yet fine-grained instance-level spatial control remains an elusive challenge, hindering real-world applications. This limitation stems from two factors: a scarcity of scalable, position-annotated datasets, and the entanglement of identity and layout by global attention mechanisms. To this end, we introduce PositionIC, a unified framework for high-fidelity, spatially controllable multi-subject customization. First, we present BMPDS, the first automatic data-synthesis pipeline for position-annotated multi-subject datasets, effectively providing crucial spatial supervision. Second, we design a lightweight, layout-aware diffusion framework that integrates a novel visibility-aware attention mechanism. This mechanism explicitly models spatial relationships via an NeRF-inspired volumetric weight regulation to effectively decouple instance-level spatial embeddings from semantic identity features, enabling precise, occlusion-aware placement of multiple subjects. Extensive experiments demonstrate PositionIC achieves state-of-the-art performance on public benchmarks, setting new records for spatial precision and identity consistency. Our work represents a significant step towards truly controllable, high-fidelity image customization in multi-entity scenarios.Code and data: https://github.com/MeiGen-AI/PositionIC.

Most event-based algorithms typically split the event stream into fixed groups (e.g., fixed time or fixed count) for downstream processing, lacking adaptivity to scene dynamics. Several adaptive partitioning strategies have been proposed, but they are unable to cope well with heterogeneous velocity scenarios (HVS) involving both fast- and slow-moving objects. To address this issue, we propose Adaptive Spatial-Temporal Window (ASTW) strategy, which simultaneously achieves temporal adaptivity and spatial locality in event partitioning. Based on the principle of maximum entropy, we derive a patch-level time window determination criterion and efficiently implement it based on event density and vectorized calculations. Experiments on publicly available event-based object detection and tracking datasets demonstrate that ASTW significantly outperforms existing state-of-the-art partitioning strategies. We also construct HetVel, the first RGB-event dual-modality dataset for HVS, and further highlight the advantages of ASTW on this challenging benchmark. We believe that our ASTW strategy and the constructed HetVel dataset will advance the field of neuromorphic vision.

Text-guided human body animation has advanced rapidly, yet facial animation lags due to the scarcity of well-annotated, text-paired facial corpora. To close this gap, we leverage foundation generative models to synthesize a large, balanced corpus of facial behavior. We design prompts suite covering emotions and head motions, generate about 80 hours of facial videos with multiple generators, and fit per-frame 3D facial parameters, yielding large-scale (prompt and parameter) pairs for training. Building on this dataset, we probe language models for bidirectional competence over facial motion via two complementary tasks: (1) Motion2Language: given a sequence of 3D facial parameters, the model produces natural-language descriptions capturing content, style, and dynamics; and (2) Language2Motion: given a prompt, the model synthesizes the corresponding sequence of 3D facial parameters via quantized motion tokens for downstream animation. Extensive experiments show that in this setting language models can both interpret and synthesize facial motion with strong generalization. To best of our knowledge, this is the first work to cast facial-parameter modeling as a language problem, establishing a unified path for text-conditioned facial animation and motion understanding.

Out-of-distribution (OOD) detection seeks to identify samples from unknown classes, a critical capability for deploying machine learning models in open-world scenarios. Recent research has demonstrated that Vision-Language Models (VLMs) can effectively leverage their multi-modal representations for OOD detection. However, current methods often incorporate intra-modal distance during OOD detection, such as comparing negative texts with ID labels or comparing test images with image proxies. This design paradigm creates an inherent inconsistency against the inter-modal distance that CLIP-like VLMs are optimized for, potentially leading to suboptimal performance. To address this limitation, we propose InterNeg, a simple yet effective framework that systematically utilizes consistent inter-modal distance enhancement from textual and visual perspectives. From the textual perspective, we devise an inter-modal criterion for selecting negative texts. From the visual perspective, we dynamically identify high-confidence OOD images and invert them into the textual space, generating extra negative text embeddings guided by inter-modal distance. Extensive experiments across multiple benchmarks demonstrate the superiority of our approach. Notably, our InterNeg achieves state-of-the-art performance compared to existing works, with a 3.47% reduction in FPR95 on the large-scale ImageNet benchmark and a 5.50% improvement in AUROC on the challenging Near-OOD benchmark.

Face relighting aims to synthesize realistic portraits under novel illumination while preserving identity and geometry. However, progress remains constrained by the limited availability of large-scale, physically consistent illumination data. To address this, we introduce POLAR, a large-scale and physically calibrated One-Light-at-a-Time (OLAT) dataset containing over 200 subjects captured under 156 lighting directions, multiple views, and diverse expressions. Building upon POLAR, we develop a flow-based generative model POLARNet that predicts per-light OLAT responses from a single portrait, capturing fine-grained and direction-aware illumination effects while preserving facial identity. Unlike diffusion or background-conditioned methods that rely on statistical or contextual cues, our formulation models illumination as a continuous, physically interpretable transformation between lighting states, enabling scalable and controllable relighting. Together, POLAR and POLARNet form a unified illumination learning framework that links real data, generative synthesis, and physically grounded relighting, establishing a self-sustaining "chicken-and-egg" cycle for scalable and reproducible portrait illumination.

We introduce Clothe and Pose, an image generation and editing task that enables users to try on garments while simultaneously adopting any desired pose. Our method takes a single user image, a set of garment images, and a reference pose as input, and outputs the user wearing the target garment in the specified pose. We also propose an evaluation framework for clothing and re-posing tasks. Our study encompasses a wide variety of garments--including athletic wear, bottoms, dresses, innerwear, and swimwear--across diverse poses. Finally, we demonstrate the utility of Clothe and Pose for human-centric editing on both real-world captures and synthetic imagery.