paperlist.ai

AI conference papers

AI conference paper library

Search papers from 30+ top conferences, including NeurIPS, ICLR, ICML, CVPR, ACL and AAAI. Filter by keyword, then open any paper for its abstract, authors and links.

ECCV 2026

DiffGI: Differentiable Geometry Images for High-Fidelity Thin-Shell 3D Generation

Eungjune Shim, Hansol Lee, Eunjung Ju

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. • •

ECCV 2026

Reward Modeling for Computer-Using Agent from Video Execution

Linxin Song, Jieyu Zhang, Huanxin Sheng, Taiwei Shi, Rahul Gupta, Yang Liu, Ranjay Krishna, Jian Kang, Jieyu Zhao

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

ECCV 2026

Learning to Compose: Revisiting Proxy Task Design for Zero-Shot Composed Image Retrieval

Jingjing Zhang, Lei Zhang, Zheren Fu, Zhendong Mao

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.

ECCV 2026

EgoPHI: Estimating 3D Hand-Object Contact and Force from Egocentric Vision

Andela Ilic, Rachel Schuchert, Yijing Jiang, Christian Holz

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

ECCV 2026

LoRC: Detecting AI-Generated Images via Low-Rank Collapse in the Semantic-Residual Space

Haozhen Yan, Ruoxin Chen, Jiahui Zhan, Taiping Yao, Bo Wang, Youchang xiao, Shouhong Ding, Liqing Zhang, Jianfu Zhang

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.

ECCV 2026

Why Can't I Open My Drawer? Mitigating Object-Driven Shortcuts in Zero-Shot Compositional Action Recognition

Geo Ahn, Inwoong Lee, Taeoh Kim, Minho Shim, Dongyoon Wee, Jinwoo Choi

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.

ECCV 2026

CrossFeat: Bridging Imaging Modalities in Feature Descriptor Space

Paul Schneider, Nazim Haouchine

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

ECCV 2026

SwiftWA: An Efficient Action-Centered World-Action Model

Chaojun Ni, Xinyu Zhou, Yukun Zhou, Jingyu Liu, Xiaofeng Wang, Zheng Zhu, Yang Wang, Qiuping Deng, Yun Ye, Hao Li, Zhichao Liu, Jindi Lv, Boyuan Wang, Guosheng Zhao, Guan Huang, Min Cao, Wenjun Mei

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.

ECCV 2026

Fixed Reality, Diffused Possibility: Disentangling Stochastic and Deterministic Latent for Cluttered Grasping

Hritam Basak, Zhaozheng Yin

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.

ECCV 2026

Exo2EgoPolicy: Pose-Aligned Cross-View Policy Learning

Hritam Basak, Hadi Tabatabaee, Xin Yang, Shreekant Gayaka, Nan Qiao, Yuyin Sun, Cheng-Hao Kuo, Zhaozheng Yin, Min Sun

Learning robot manipulation policies from large-scale human videos is challenging due to the viewpoint mismatch between thirdperson observations and egocentric robot control. To this end, we introduce Exo2EgoPolicy, a geometry-aware framework that learns egocentric manipulation policies from monocular exocentric video via viewpointaligned pose translation. Unlike prior Exo→Ego approaches that rely on diffusion-based pixel synthesis and often produce temporally inconsistent predictions, our method operates directly on a structured pose manifold and explicitly models the geometric ambiguity inherent in monocular observation. We formulate exocentric-to-egocentric translation as a viewpoint-conditioned latent variable model on the SE(3) manifold that disentangles pose from camera transformation and captures the ambiguity induced by unknown camera extrinsics. Under bounded viewpoint motion, the formulation yields pose trajectories that are identifiable up to a global rigid transformation and supports stable sequential inference through temporal regularization. Beyond representation alignment, we show that for quasi-static manipulation tasks whose rewards depend primarily on relative hand–object geometry, pose captures the key task-relevant information required for policy learning. This enables policy co-training on pose-translated human demonstrations alongside limited robot teleoperation data within a unified human–robot state space. Empirically, Exo2EgoPolicy improves temporal consistency, cross-view alignment, and out-of-distribution policy transfer compared to pixelbased translation and other cross-view co-training baselines, achieving 20–30% higher task success on manipulation benchmarks. These results suggest that geometry-aligned pose representations provide a scalable foundation for cross-view policy learning from human video. Website

ECCV 2026

JanusMesh: Fast and Zero-Shot 3D Visual Illusion Generation via Cross-Space Denoising

Siang Ling Zhang, Huai-Hsun Cheng, Tsung-Ju Yang, Yu-Lun Liu

Creating 3D visual illusions, a single 3D mesh that revealsentirely different semantics from various viewing angles, is a fascinat-ing but tough challenge. Existing optimization-based methods are slowand can produce oversaturated colors. In contrast, naive stitching ap-proaches fail to produce geometrically coherent objects. This results invisible unnatural seams and semantic leaks. In this paper, we presenta fast and training-free framework for generating text-driven 3D visualillusions. Our approach decouples the generation into two stages. First,we propose a cross-space dual-branch denoising process. This processdynamically decodes 3D latents into voxel space for CLIP-guided orien-tation alignment and Signed Distance Field (SDF) blending, which en-sures seamless geometric fusion. Second, we introduce a view-conditionedtexture synthesis module that projects and aggregates view-specific 2Ddiffusion priors onto the fused geometry. Extensive experiments demon-strate that our method generates highly realistic, dual-semantic 3D illu-sions in just 3–5 minutes. It significantly outperforms existing methodsin geometric integrity, semantic recognizability, and efficiency. Projectpage: https://siang1105.github.io/JanusMesh.github.io/Target Junction view Target Junction view Target Junction viewview view view“A parrot”𝜃" 𝜃" 𝜃"Target Target Targetview view view𝜃! 𝜃! 𝜃!“A boat”❌ Severe over-saturation ❌ Visible backside leakage ✅ Dual-semantic coherency❌ Slow optimization (~40 min) ❌ Unnatural geometric seams ✅ Fast & Training-free (~3-5 min)(a) SDS-Based Method (b) Direct Concatenation (c) OursFig. 2: Comparison of 3D visual illusion generation methods. (a) SDS-BasedMethods suffer from severe over-saturation and slow optimization. (b) Direct Con-catenation exposes unnatural geometric seams and semantic leakage at target views(red arrows). (c) Our method creates a seamless, dual-semantic coherent 3D mesh.Unlike previous approaches, our method does not require training. It generates high-quality 3D visual illusions in just 3–5 minutes while completely preventing geometricinterference between the two semantics.

ECCV 2026

TaxoMIL: Taxonomy-Constrained Learning for Hierarchical Whole Slide Image Analysis

Chaeyeon Lee, Khang Quoc, Jinsol Song, Yosep Chong, Kwangil Yim, Jin Tae Kwak

Whole slide image (WSI) analysis is central to computationalpathology, with multiple instance learning (MIL) emerging as the stan-dard pipeline for slide-level diagnosis. However, conventional approachesformulate WSI diagnosis as a flat classification task over discrete labels,contradicting the inherently hierarchical, coarse-to-fine nature of clinicalreasoning. Although recent hierarchical classifiers and vision–languagemodels (VLMs) have sought to address this structural gap, they eitherfail to capture semantic continuity between related diagnoses or sufferfrom unconstrained text generation that produces taxonomic hallucina-tions and parent–child label violations. To address these limitations, wepropose TaxoMIL, a taxonomy-constrained framework that reformulatesWSI diagnosis as a multi-granularity text generation task. TaxoMIL uti-lizes a dual-head Transformer decoder to generate coarse- and fine-leveldiagnostic text, and introduces taxonomy-guided objectives that explic-itly structure the label embedding space and strictly ground slide-level vi-sual representations within the clinical taxonomy. Extensive experimentsacross three diverse WSI datasets demonstrate that TaxoMIL consis-tently outperforms state-of-the-art MIL classifiers and VLM-based gen-erative methods, yielding accurate and hierarchy-aware diagnostic pre-dictions. The code is released at https://github.com/QuIIL/TaxoMIL..

ECCV 2026

OmniScript: Towards Audio-Visual Script Generation for Long-Form Cinematic Video

Junfu Pu, Yuxin Chen, Teng Wang, Ying Shan

Current multimodal large language models (MLLMs) havedemonstrated remarkable capabilities in short-form video understand-ing, yet translating long-form cinematic videos into detailed, temporallygrounded scripts remains a significant challenge. This paper introducesthe novel video-to-script (V2S) task, aiming to generate hierarchical,scene-by-scene scripts encompassing character actions, dialogues, expres-sions, and audio cues. To facilitate this, we construct a first-of-its-kindhuman-annotated benchmark and propose a temporally-aware hierarchi-cal evaluation framework. Furthermore, we present OmniScript, an 8B-parameter omni-modal (audio-visual) language model tailored for long-form narrative comprehension. OmniScript is trained via a progressivepipeline that leverages chain-of-thought supervised fine-tuning for plotand character reasoning, followed by reinforcement learning using tem-porally segmented rewards. Extensive experiments demonstrate that de-spite its parameter efficiency, OmniScript significantly outperforms largeropen-source models and achieves performance comparable to state-of-the-art proprietary models, including Gemini 3-Pro, in both temporallocalization and multi-field semantic accuracy. The code is available athttps://github.com/TencentARC/OmniScript.

ECCV 2026

FUSE: A Flow-based Mapping Between Shapes

Lorenzo Olearo, Giulio Viganò, Daniele Baieri, Filippo Maggioli, Simone Melzi

We introduce a novel neural representation for maps between3D shapes based on flow-matching models, which is computationally ef-ficient and supports cross-representation shape matching without large-scale training or data-driven procedures. 3D shapes are represented asthe probability distribution induced by a continuous and invertible flowmapping from a fixed anchor distribution. Given a source and a targetshape, the composition of the inverse flow (source to anchor) with theforward flow (anchor to target), we map points between the two surfaces.By encoding the shapes with a pointwise task-tailored embedding, thisconstruction provides an invertible and modality-agnostic representationof maps between shapes across point clouds, meshes, signed distancefields (SDFs), and volumetric data. The resulting representation con-sistently achieves high coverage and accuracy across diverse benchmarksand challenging settings in shape matching. Beyond shape matching, ourframework shows promising results in other tasks, including UV mappingand registration of raw point cloud scans of human bodies.

ECCV 2026

Decompose, Compare, and Decide: Multimodal LLMs are Implicit Few-Shot Learners

Yunhan Wang, Eshika Khandelwal, Edson Araujo, Walid Bousselham, Nina Shvetsova, Hilde Kuehne

Multimodal Large Language Models (MLLMs) have demon-strated remarkable abilities when analyzing images, yet translating thesecapabilities to few-shot image classification remains challenging. To bridgethis gap, we present DeCoDe, a simple yet effective technique that en-ables off-the-shelf MLLMs to act as strong few-shot classifiers withoutany additional training. Our approach builds on the idea of few-shotclassification as a set of pairwise image comparisons, decomposing thetask into a set of binary decisions. Given a query image and a supportimage from a candidate class, the MLLM is prompted to decide whetherthe two images depict the same class. The logit corresponding to an af-firmative response is then used as a similarity score to assign the queryimage to the most likely class. While this already yields good results,we show that providing additional high-level information, such as thedata domain, to the model further improves performance. Our evalu-ation provides an extensive analysis of various inference variants on asuite of twelve datasets, six established and six newly curated few-shotbenchmarks spanning across diverse domains. The results show that theproposed simple decomposition technique can turn off-the-shelf MLLMsinto powerful few-shot learners, significantly outperforming current state-of-the-art few-shot methods on both standard and novel domains. Codeis available at https://github.com/yunhanwang1105/DeCoDe.

ECCV 2026

CoSPlan: Corrective Sequential Planning via Scene Graph Incremental Updates

Shresth Grover, Priyank Pathak, Akash Kumar, Yogesh Rawat

Vision Language Models (VLMs) have shown promising plan-ning capabilities, yet their success remains confined to the text domain,leaving visual decision-making relatively underexplored. Addressing thisgap, we introduce Corrective Sequence Planning (CoSPlan) bench-mark, where VLMs must plan a sequence of visual actions from an initialscene to a target scene. CoSPlan evaluates models on their ability to imag-ine and execute a coherent set of visual steps required to reach the goal(Step Completion). To prevent any shortcuts that simply describethe final scene, we introduce an erroneous action in decision making,which must be detected (Error Detection) and corrected to reach thegoal, enabling a deeper understanding of the task. CoSPlan spans across4 tasks: maze navigation, block re-arrangement, image reconstruction,and object re-organization. Despite using advanced reasoning strategiessuch as Chain-of-Thought and Scene Graphs, VLMs struggle on CoSPlan,while still showing promising performance in the text domain. Addressingthis, we propose Scene Graph Incremental updates (SGI), a noveltraining-free method to transform images into ‘textual’ scene graphs, en-abling step-by-step reasoning through iterative scene graph refinement.SGI yields an average of ≃ 4.4% ↑ on CoSPlan w/ generalization onPlanBench and VQA. Link for solving puzzles on the project page.

ECCV 2026

Evaluating the Interpretability of Sparse Autoencoders with Concept Annotations

Jonas Klotz, Cassio F. Dantas, Pallavi Jain, Diego Marcos, Begum Demir

Sparse autoencoders (SAEs) are increasingly used to extract interpretable concepts from vision and vision language models, yet existing evaluation methods largely rely on proxy metrics or qualitative inspection rather than measuring semantic correspondence. We present a human-grounded evaluation framework that quantifies alignment between SAE latents and human-annotated concepts, without requiring user studies, and validate this matching through targeted attribute perturbations. To enable this intervention-style evaluation in vision, we construct synCUB and synCOCO, synthetic benchmarks of paired images that differ in exactly one attribute. We introduce Fully-Binary Matching Pursuit (FBMP), a coalition-based matching procedure that supports many-to-one mappings between SAE latents and annotated concepts, and consistently outperforms one-to-one baselines. For functional validation, we propose a Targeted Attribute Perturbation Alignment Score (TAPAScore), which tests whether matched concepts respond selectively and in the expected direction under targeted image-level attribute perturbations. Under sanity checks, our matching and TAPAScore are the only evaluated metrics that reliably distinguish trained SAEs from untrained ones. Across SAEs trained on CLIP and DINOv2 embeddings, we find that increased overcompleteness can reduce perturbation alignment, indicating a reduction in interpretability. Our evaluation framework suggests that moderate dictionary sizes provide the best trade-off, yielding the most interpretable SAEs. Code and datasets are available at https://github.com/JonasKlotz/sae-concept-eval.

ECCV 2026

Indelible Backdoors: On the Limits of Post-Training Defenses

Jingyi Guo, Thuy Dung Nguyen, Taylor T Johnson, Kevin Leach

Deep neural networks remain vulnerable to backdoor at-tacks, in which adversaries embed hidden triggers during training tocause misclassification of any trigger-carrying input into an attacker-chosen target class, while maintaining normal behavior on clean inputs.Though recent fine-tuning-based defenses have shown promise in miti-gating backdoor attacks and are particularly practical in continual learn-ing and MLaaS scenarios, where backdoors are removed from pretrainedmodels using small clean datasets, we demonstrate that this defensiveparadigm remains fundamentally exploitable by an adaptive adversary.We study a threat model in which attackers anticipate fine-tuning-baseddefenses and craft backdoors whose resulting optimization signals areentangled with those of the clean task. Our approach leverages FisherInformation to identify task-critical neurons and optimizes a trigger gen-erator such that backdoor gradients align with benign gradients on theseneurons, ensuring that any attempt to suppress the backdoor inevitablyinterferes with clean task preservation. As a result, fine-tuning defensesfails to effectively reduce attack success rate without incurring substan-tial degradation in clean accuracy. Through extensive evaluation againstseven attacks and eight defenses, we show our method consistently showsa stronger effect than existing backdoor attacks across all settings. Ourwork reveals a critical vulnerability in fine-tuning-based defenses andhighlights the need for more robust defense mechanisms against adap-tive adversaries.

ECCV 2026

Control-DINO: Feature Space Conditioning for Controllable Video Diffusion

Edoardo Dominici, Thomas Deixelberger, Konstantinos Vardis, Markus Steinberger

Video diffusion models have recently been applied with success to problems in content generation, novel view synthesis, and, more broadly, world simulation. Many applications in generation and transfer rely on conditioning these models, typically through perceptual, geometric, or simple semantic signals, fundamentally using them as generative renderers. At the same time, high-dimensional features obtained from large-scale self-supervised learning on images or point clouds are increasingly used as a general-purpose interface for vision models. The connection between the two has been explored for subject specific editing, aligning and training video diffusion models, but not in the role of a dense conditioning signal for pretrained video diffusion models. Features obtained through self-supervised learning like DINOv3, contain a lot of entangled information about style, lighting and semantics of the scene. This makes them great at reconstruction tasks but limits their generative capabilities. In this paper, we show how we can use the features for tasks such as video domain transfer and video-from-3D generation. We introduce a lightweight control architecture and training strategy that decouples appearance from other features that we wish to preserve, enabling robust control for appearance changes such as stylization and relighting. Furthermore, we show that low spatial resolution can be compensated by higher feature dimensionality, improving controllability in generative rendering from explicit spatial representations.

ECCV 2026

P-CORE: Self-Supervised Surface Consistency for Point-Based Neural Editing

Yanshu Zhang, Shichong Peng, Mehran Aghabozorgi, Alireza Moazeni, Ke Li

Advances in neural rendering have enabled high-fidelity multi-view reconstruction of 3D scenes. However, free-form non-rigid shapeediting remains a significant challenge. Point-based neural representationsare highly desirable for multi-view reconstruction because they lack fixedconnectivity, which does not constrain the learned surface topology tothat of the initialization. Yet this same property causes point-based rep-resentations to struggle with holes and surface discontinuities under largedeformations. To address this, we propose a novel self-supervised methodto enable point-based representations to adapt to large deformations with-out requiring ground truth multi-view images of deformed geometry. Thekey idea is to generate random deformations and to ensure consistencyin the predicted surface before and after deformation. In particular, thesurface prediction from the deformed point cloud should be the sameas the deformation applied to the surface prediction from the originalpoint cloud. We incorporate our approach into attention-based pointrepresentations, which differ from splatting-based point representationsin their use of a learned interpolation kernel between points as opposedto a Gaussian kernel around each point. This learned interpolation kernelcan learn to adapt to large deformations, without requiring addition orremoval of points. We show that our framework significantly enhancesits robustness to large deformations. Experiments on synthetic geometryediting benchmarks (Neural Editor, Objaverse) demonstrate that ourapproach outperforms existing point-based methods in zero-shot editingand significantly reduces artifacts. Furthermore, qualitative results on theDTU and Mip-NeRF 360 datasets demonstrate our method’s effectivenesson real-world scenes.

Keep browsing

More papers from top AI conferences

30+ top conferences, by conference, year, and topic.