ECCV 2026 Papers

3D Morphable Models (3DMMs) remain the standard parametric shape priors for many state-of-the-art 3D face reconstruction algorithms. However, as these models are derived from a finite number of 3D face samples, they inherit the morphological biases of their training data, potentially limiting their generalizability across diverse global populations. In this paper, we propose a novel framework to analyze 3DMM reconstructions through the lens of surface curvature, with the objective to discover, quantify and visualize biases. While standard evaluation metrics often rely on Euclidean distances, our reconstruction error captures subtle surface nuances such as local topology or undulations. To do so, we leverage the Laplace-Beltrami Operator (LBO) to generate high-resolution curvature error maps, providing a localized and geometrically meaningful visualization of discrepancies between ground truth faces and reconstructed meshes. We derive from it an error metric that we validated through a user study, observing a significantly higher correlation to human perception compared to traditional methods. Furthermore, we conduct extensive experiments across several 3DMM bases and fitting algorithms, uncovering systematic age-related biases and providing preliminary evidence of biases associated with gender and ethnicity. Our findings highlight the necessity of adopting curvature-aware evaluation protocols to ensure demographic fairness and geometric precision in future 3D face reconstruction research. The code and annotation data are available at https://github.com/artefactory/3dface-fairness.

Training-free image editing with diffusion models is highlydesirable yet is complex and remains a significant challenge. While recentoptimisation-based methods achieve strong zero-shot edits from text,they still struggle to preserve identity and capture intricate details, suchas facial structure, surface texture, or object-specific geometry, that ex-ist below the level of linguistic abstraction. To address this fundamentalgap, we propose Concept Distillation Sampling (CDS). To the best of ourknowledge, we are the first to introduce a unified, training-free frameworkfor target-less, multi-concept image editing.CDS overcomes this linguistic bottleneck of previous methods by anchor-ing the editing process in the certainty of pretrained LoRA adapters.We integrate a highly stable distillation backbone (featuring orderedtimesteps, regularisation, and negative-prompt guidance) with a noveldynamic weighting mechanism. This approach enables the compositionand control of multiple visual concepts directly within the diffusion pro-cess, utilising spatially-aware priors from pretrained LoRA adapters with-out causing concept clashing. Our method preserves instance conceptidentity without requiring reference samples of the desired edit. Extensivequantitative and qualitative evaluations demonstrate that CDS estab-lishes a new state-of-the-art over existing training-free editing and multi-LoRA composition methods on the InstructPix2Pix and ComposLoRAbenchmarks. Project Page: https://nickyfot.github.io/cds/.

This paper introduces the concept of Microscopic Spatial In-telligence (MiSI), the capability to perceive and reason about the spatialrelationships of invisible microscopic entities, which is fundamental toscientific discovery. To assess the potential of Vision-Language Models(VLMs) in this domain, we propose a systematic benchmark frameworkMiSI-Bench. This framework features over 163,000 question-answer pairsand 587,975 images derived from approximately 4,000 molecular struc-tures, covering nine complementary tasks that evaluate abilities rangingfrom elementary spatial transformations to complex relational identifi-cations. Experimental results reveal that current state-of-the-art VLMsperform significantly below human level on this benchmark. However,a fine-tuned 7B model demonstrates substantial potential, even sur-passing humans in spatial transformation tasks, while its poor perfor-mance in scientifically-grounded tasks like hydrogen bond recognitionunderscores the necessity of integrating explicit domain knowledge forprogress toward scientific AGI. The datasets are available at https://huggingface.co/datasets/zongzhao/MiSI-bench.

With the growing demand for immersive visual experiences, high-quality omnidirectional images (ODIs) have become increasingly important. However, limitations in imaging devices and transmission bandwidth often lead to low-resolution ODIs, hindering the rendering of fine-grained 360◦ details, especially in the presence of real-world degradations and geometric distortions. Existing real-world super-resolution (Real-SR) methods are inadequate for ODIs, as their degradation models fail to account for the complex imaging pipeline involving fisheye capture and Equirectangular Projection (ERP), introducing severe aliasing and projection-specific distortions. To address these challenges, we propose D2R2OSR, a Degradation-Disentangled Representation framework for Real-world Omnidirectional image Super-Resolution. D2R2OSR explicitly models degradations arising from both fisheye imaging and ERP projection, guided by two key insights: (1) projection priors play a critical role in shaping real-world degradations, and (2) human perception in immersive environments is inherently viewpoint-centric. Accordingly, we introduce a Perspective Projection Representation (PPR) operating alongside the ERP branch to capture viewpoint-aware features, together with a Degradation-Specific Module (DSM) that jointly models ERPinduced geometric distortions and PPR-specific real-world degradations. Extensive experiments demonstrate that D2R2OSR achieves state-of-theart performance and produces visually compelling, high-fidelity omnidirectional Real-SR results while maintaining favorable computational efficiency for low-resource deployment.

Bird’s-Eye View (BEV) representations derived from multi-camera input have become a central interface for online HD map con-struction. However, most approaches rely solely on ego-centric super-vision, requiring large-scale scene structure to be inferred from incom-plete observations, occlusions, and diminishing information density atlong range, where perspective effects and spatial sparsity hinder consis-tent structural reasoning. We introduce Cross-View Supervision (CVS),a representation learning paradigm that transfers geometric and topolog-ical priors from an ego-aligned overhead perspective into camera-basedBEV encoders. Rather than adding auxiliary semantic losses, CVS alignsrepresentations in a shared BEV feature space and distills globally con-sistent structural knowledge from a perspective-privileged teacher intothe ego-centric backbone. This supervision enhances structural coher-ence without modifying the inference architecture or requiring overheadinput at test time. Experiments on nuScenes using ego-aligned aerialimagery from the AID4AD cross-view extension demonstrate consistentimprovements over StreamMapNet while maintaining identical camera-only inference. CVS yields +3.9 mAP in the standard 60 × 30 m regionand +9.9 mAP in the extended 100 × 50 m setting, corresponding toa 44% relative gain at long range. These results highlight perspective-privileged structural supervision as a promising training principle forimproving BEV representation learning in HD map construction. Theproject repository is available athttps://github.com/DriverlessMobility/CrossViewSupervision.

Image relighting modifies illumination while preserving non-lighting content such as identity and geometry. Existing diffusion-basedmethods often suffer from unstable illumination changes or inconsistentcontent preservation under complex lighting, as they lack an explicitmechanism to learn feature transformations between images. We refor-mulate relighting as an illumination feature transport problem and in-troduce Consistent Feature Transport (CFT), a training principle thatexplicitly enforces illumination-consistent transport between source andtarget image distributions. Built upon rectified flow, CFT jointly modelsnoise-to-image generation and illumination-consistent source-to-targettransport through trajectory-level supervision. This dual-transport for-mulation encourages isolation of illumination-specific variations whilepreserving content-aligned features. To support complex lighting scenar-ios, we construct a large-scale portrait relighting dataset with diverserelighting effects. Experiments show consistent improvements over ex-isting state-of-the-art relighting approaches and demonstrate that CFTcan generalize to other editing tasks, including style transfer. Code isavailable at https://github.com/Dixin-Lab/CFT.

Interleaved text-and-image generation represents a signifi-cant frontier for Multimodal Large Language Models (MLLMs), offeringa more intuitive way to convey complex information. Current paradigmsrely on either image generation or retrieval augmentation, yet they typi-cally treat the two as mutually exclusive paths, failing to unify factualitywith creativity. We argue that the next milestone in this field is AgenticTool Planning, where the model serves as a central controller that au-tonomously determines when, where, and which tools to invoke to pro-duce interleaved responses for visual-critical queries. To systematicallyevaluate this paradigm, we introduce ATP-Bench, a novel benchmarkcomprising 7,702 QA pairs (including 1,592 VQA pairs) across eight cat-egories and 25 visual-critical intents, featuring human-verified queriesand ground truths. Furthermore, to evaluate agentic planning indepen-dent of end-to-end execution and changing tool backends, we proposea Multi-Agent MLLM-as-a-Judge (MAM) system. MAM evaluates tool-call precision, identifies missed opportunities for tool use, and assessesoverall response quality without requiring ground-truth references. Ourextensive experiments on 10 state-of-the-art MLLMs reveal that modelsstruggle with coherent interleaved planning and exhibit significant varia-tions in tool-use behavior, highlighting substantial room for improvementand providing actionable guidance for advancing interleaved generation.

Multimodal Large Language Models (MLLMs) are increasingly applied to pixel-level vision tasks, yet their intrinsic capacity for spatial understanding remains poorly understood. We investigate segmentation capacity through a layerwise linear probing evaluation across the entire MLLM pipeline: vision encoder, adapter, and LLM. We further conduct an intervention based attention knockout analysis to test whether cross-token attention progressively refines visual representations, and an evaluation of bidirectional attention among image tokens on spatial consistency. Our analysis reveals that the adapter introduces a segmentation representation drop-o!, but LLM layers progressively recover through attention-mediated refinement, where correctly classified tokens steer misclassified neighbors toward the correct label. At early image token positions, this recovery is bounded by causal attention, which bidirectional attention among image tokens alleviates. These findings provide a mechanistic account of how MLLMs process visual information for segmentation, informing the design of future segmentation-capable models.

Interpreting the decisions of complex computer vision modelsis crucial to establish trust and accountability, especially in safety-criticaldomains. An established approach to interpretability is generating visualattribution maps that highlight regions of the input most relevant to themodel’s prediction. However, existing methods face a three-way trade-off. Propagation-based approaches are efficient, but they can be biasedand architecture-specific. Meanwhile, perturbation-based methods arecausally grounded, yet they are expensive and for vision transformersoften yield coarse, patch-level explanations. Learning-based explainersare fast but usually optimize surrogate objectives or distill from heuris-tic teachers. We propose a learning scheme that instead optimizes dele-tion and insertion metrics directly. Since these metrics depend on non-differentiable sorting and ranking, we frame them as permutation learn-ing and replace the hard sorting with a differentiable relaxation usingGumbel-Sinkhorn. This enables end-to-end training through attribution-guided perturbations of the target model. During inference, our methodproduces dense, pixel-level attributions in a single forward pass with op-tional, few-step gradient refinement. Our experiments demonstrate con-sistent quantitative improvements and sharper, boundary-aligned expla-nations, particularly for transformer-based vision models. Code and pre-trained models are available at https://github.com/dschinagl/AHA.

Multimodal large language models (MLLMs) ground wholeobjects well from free-form language queries, but they struggle whenthe query names a part rather than the object. We trace this to amissing object-part hierarchy, since parts are localized in the same sin-gle step used for objects. We propose Object-Part Hierarchical Reflec-tive Grounding (OP-HRG), a coarse-to-fine reasoning-guided groundingstrategy that first localizes the parent object and then the part within it.A self-check then reflects on the result, with an extension to re-encodethe predicted crop to inspect the region it is correcting. We introducea part-aware GRPO framework to train our pipeline with stage-wise re-wards. A 4B model trained this way outperforms 7B grounding LLMs andSAM3 across PascalPart, PartImageNet, and InstructPart, and transfersto reasoning segmentation. 1

Existing 3D generative models predominantly rely on implicit volumetric representations, which inherently enforce watertight topology and struggle to faithfully represent thin-shell and non-manifold geometries such as garments. While geometry image-based approaches o(cid:27)er a surface-centric alternative, existing methods typically rely on discrete binary occupancy maps whose resolution-dependent boundary encoding causes staircase artifacts and information loss upon downsampling, while surface reconstruction remains a non-di(cid:27)erentiable postprocessing step disconnected from the learning pipeline. 1 To address this, we propose Di(cid:27)erentiable Geometry Image (Di(cid:27)GI) , an end-to-end 3D-to-2D mapping framework that seamlessly integrates surface representation and geometric optimization. Di(cid:27)GI replaces conventional binary maps with a continuous 2D Truncated Signed Distance Function (TSDF), which encodes boundary position at subpixel precision within a (cid:28)xed grid resolution, e(cid:27)ectively eliminating resolutiondependent staircase artifacts even under aggressive downsampling. Building on this continuous (cid:28)eld, we introduce a di(cid:27)erentiable Marching Squares algorithm based on analytical linear interpolation, allowing gradients from 3D surface losses to propagate seamlessly back to the 2D latent space. Leveraging this di(cid:27)erentiable pipeline, we train a Di(cid:27)GI-VAE augmented with a geometry-aware normal rendering loss to compress complex 3D surfaces into an ultra-compact 32 × 32 latent space. Finally, we instantiate a transformer-based latent di(cid:27)usion model on top of this space for conditional 3D generation, showing that the proposed representation readily supports e(cid:30)cient generative modeling. Extensive experiments on garment and object datasets demonstrate that our method achieves superior reconstruction (cid:28)delity and boundary precision compared to prior geometry-image and voxel-based approaches, while requiring signi(cid:28)cantly fewer computational resources. • •

Computer-use agents (CUAs) are becoming increasingly ca-pable; however, it remains difficult to scale evaluation of whether a tra-jectory truly fulfills a user instruction. In this work, we study rewardmodeling from execution video: a sequence of keyframes from an agenttrajectory that is independent of the agent’s internal reasoning or ac-tions. Although video-execution modeling is method-agnostic, it presentskey challenges, including highly redundant layouts and subtle, localizedcues that determine success. We introduce Execution Video Reward 53k(ExeVR-53k), a dataset of 53k high-quality video–task–reward triplets.We further propose adversarial instruction translation to synthesize neg-ative samples with step-level annotations. To enable learning from long,high-resolution execution videos, we design spatiotemporal token prun-ing, which removes homogeneous regions and persistent tokens while pre-serving decisive UI changes. Building on these components, we fine-tunean Execution Video Reward Model (ExeVRM) that takes only a user in-struction and a video-execution sequence to predict task success. Our Ex-eVRM 8B achieves 84.7% accuracy and 87.7% recall on video-executionassessment, outperforming strong proprietary models such as GPT-5.2and Gemini-3 Pro across Ubuntu, macOS, Windows, and Android, whileproviding more precise temporal attribution. These results show thatvideo-execution reward modeling can serve as a scalable, model-agnosticevaluator for CUAs.Code: https://github.com/limenlp/ExeVRMModel: https://huggingface.co/lime-nlp/ExeVRM-8BDataset: https://huggingface.co/datasets/lime-nlp/ExeVR-53K

Composed Image Retrieval (CIR) retrieves a target imagefrom a reference image and a textual modification. While supervisedCIR relies on costly triplets, Zero-Shot CIR (ZS-CIR) alleviates thisreliance through proxy tasks trained on image–text pairs. However, ex-isting proxy tasks primarily enhance visual and textual representationsto accommodate a predefined composition mechanism such as pseudo-word injection into a frozen text encoder or linear feature arithmetic.As a result, the composition function itself remains unlearned, limitingthe model’s ability to express diverse and fine-grained semantic mod-ifications. To address this, we propose FoCo, which models composi-tion as two coordinated stages: focusing on modification-relevant visualcontent, and then completing the target semantics. We realize thesethrough two proxy tasks: text-anchored visual aggregation to selectivelygather visual content guided by localized textual semantics, and context-conditioned semantic completion to transform these aggregated visualswith the remaining scene context into a coherent composed represen-tation. The tasks are trained jointly with a cross-instance contrastiveobjective, encouraging semantic diversity and discouraging shortcut com-position strategies. Extensive experiments on four ZS-CIR benchmarksshow FoCo’s state-of-the-art performance and improved generalization.

Understanding hand-object interaction from egocentric vi-sion is essential for modeling how people physically engage with thesurrounding world. Yet reasoning about physically grounded interactionrequires estimating the forces acting on hands and objects, beyond lo-calizing contact. We present EgoPHI, the first method that jointly esti-mates dense contact maps and 3D force distributions on hand and objectmeshes from a single egocentric RGB image and object geometry. To ad-dress the lack of scalable ground-truth force annotations, we introducea physics-based simulation pipeline that augments existing hand-objectdatasets with dense per-vertex force supervision. EgoPHI then learnsdense 3D contact and force on interacting hand and articulated objectmeshes, extending vision-based force estimation beyond image-space orplanar settings. Our evaluation on in-distribution and out-of-distributionbenchmarks shows that EgoPHI improves force estimation over existingapproaches while generalizing to unseen datasets. To evaluate sim-to-realtransfer, we constructed two physical objects that capture dense objectcontact and force magnitude and used them to record a dataset of in-teractions from eight participants across diverse touch and grasp types.Our results demonstrate that EgoPHI recovers meaningful 3D contactand force distributions in simulated, out-of-distribution, and real-worldsettings, advancing egocentric hand-object understanding from contactlocalization toward physically grounded interaction reasoning.Code: https://github.com/eth-siplab/EgoPHI

Modern generators faithfully model macroscopic semantics,producing synthetic images that appear highly realistic. Consequently,decisive forensic cues reside in subtle non-semantic visual discrepancies.To reveal these cues, we revisit AIGI detection from a geometric perspec-tive and identify an architecture-agnostic signature. Specifically, mod-ern generators exhibit low-rank collapse (i.e., rank degeneracy) in thesemantic-residual orthogonal subspace while largely preserving the dom-inant semantic direction. This structural flattening consistently emergesduring the final decoding stage, forming a shared bottleneck across di-verse generator architectures. Motivated by this signature, we proposeLoRC, a framework that decouples semantic dominance to capture thecollapsed residual geometry induced by the generative decoding bottle-neck. Our method improves accuracy by an average of 7.0% across mul-tiple benchmarks and achieves 97.0% accuracy on 39 unseen generators.These results demonstrate strong cross-model generalization and robust-ness, making LoRC a reliable approach for AIGI detection in complexreal-world environments.

Zero-Shot Compositional Action Recognition (ZS-CAR) re-quires recognizing novel verb–object combinations composed of previ-ously observed primitives. In this work, we tackle a key failure mode:models predict verbs via object-driven shortcuts (i.e., relying on the la-beled object class) rather than temporal evidence. We argue that sparsecompositional supervision and verb–object learning asymmetry can pro-mote object-driven shortcut learning. Our analysis with proposed di-agnostic metrics shows that existing methods overfit to training co-occurrence patterns and underuse temporal verb cues, resulting in weakgeneralization to unseen compositions. To address object-driven short-cuts, we propose Robust COmpositional REpresentations (RCORE) withtwo components. Co-occurrence Prior Regularization (CPR) adds ex-plicit supervision for unseen compositions and regularizes the modelagainst frequent co-occurrence priors by treating them as hard nega-tives. Temporal Order Regularization for Composition (TORC) enforcestemporal-order sensitivity to learn temporally grounded verb representa-tions. Across Sth-com and EK100-com, RCORE reduces shortcut diagnos-tics and consequently improves compositional generalization. The codeis available at https://github.com/KHU-VLL/RCORE.

Most advances in keypoint descriptions address monomodalsettings, where image variations arise from viewpoint, illumination, orcontrast changes. Multimodal scenarios involve images produced by fun-damentally different sensing processes, such as multispectral imaging,RGB-depth, satellite imagery, or medical imaging, causing the samestructures to appear differently. A common solution to cross-modal de-scription is to train descriptors for each modality pair, which requires re-training whenever the modalities change, or to train large models, whichincur a significant increase in runtime. Instead, we propose CrossFeat,a framework that enables an existing monomodal descriptor to operateacross modalities. Our method learns a crossing function in descriptorspace that maps features from one modality to a representation com-patible with another. To preserve the structural information capturedby the original descriptor, CrossFeat introduces a geometry–appearancedisentanglement such that only appearance is altered while the geomet-ric properties are preserved. Experiments across multiple domains anddatasets demonstrate improved performance in multimodal matching.Code available here: https://github.com/paulschneider01/CrossFeat

World–Action Models (WAMs) initialized from pre-trainedvideo generation backbones have demonstrated remarkable potential forrobot policy learning. However, existing approaches face two critical bot-tlenecks: joint reasoning over future visual dynamics and actions incurssubstantial inference overhead, and joint modeling entangles visual andmotion representations, making motion prediction dependent on videoforecasts. To address these issues, we introduce GigaWorld-Policy, anaction-centered WAM that learns 2D pixel–action dynamics while en-abling efficient action decoding with optional video generation. Specif-ically, the model predicts future action sequences conditioned on thecurrent observation and, during training, generates future videos condi-tioned on the predicted actions and the same observation. Supervisionfrom both action prediction and video generation provides richer learn-ing signals and encourages physically plausible actions through visual-dynamics constraints. With a causal design that prevents future-videotokens from influencing action tokens, explicit future-video generationis optional at inference time, enabling faster deployment. To supportthis paradigm, we curate a diverse, large-scale robot dataset to pre-train the model. Experiments on real-world robotic platforms show thatGigaWorld-Policy runs 9× faster than the WAM baseline Motus whileimproving task success rates by 7%; compared with π0.5 , it improvesperformance by 95% on RoboTwin 2.0.

Recent diffusion-based generative models have shown strongcapabilities in 3D scene understanding and affordance reasoning. How-ever, existing methods predominantly rely on deterministic global en-coders or fully stochastic diffusion backbones, which fail to disentanglestructural certainty from contextual uncertainty, a key requirement forreliable affordance-grasp prediction in cluttered, partially observed en-vironments. We propose a split-latent hierarchical diffusion frameworkthat explicitly decomposes the global scene representation into two com-plementary subspaces: a fixed latent encoding deterministic geometricand semantic priors, and a diffused latent capturing stochastic scene-level variations through a Global Conditional Diffusion guided by RGBcontext. To propagate global uncertainty to fine-grained reasoning, weintroduce a Local Conditional Diffusion that predicts dense, point-wiseaffordance and grasp fields conditioned jointly on both latents, producinggeometrically precise yet contextually diverse predictions. Furthermore,we incorporate Harmonic Grasp Field, a differential geometric regulariza-tion that enforces smooth, obstacle-aware grasp manifolds via harmonicpotential constraints. This unified formulation bridges deterministic ge-ometry with probabilistic reasoning, yielding physically consistent anduncertainty-aware grasp generation. Extensive experiments on cluttered3D affordance and grasping benchmarks demonstrate superior stability,generalization, and diversity compared to deterministic and single-latentdiffusion approaches. Project Website.