History

CVPR-2025-reading-papers-with-code

CVPR-2026-reading-papers-with-code

收集CVPR 2026论文&源码

收集全网对CVPR 2026论文的优质讲解

---

注1：欢迎各位作者大佬提交issue，分享CVPR 2026论文和开源项目！

注2：关于CV领域顶级期刊（TPAMI、IJCV等）论文解读大盘点，详见： https://github.com/Paper2Chinese/Paper2Chinese

注3：关于人工智能领域NeurIPS顶会论文解读大盘点，详见： https://github.com/Paper2Chinese/NeurIPS2024-Reading-Paper-With-Code

【CVPR 2026 论文开源目录】

图像去雾(Dehazing)

Bilevel Layer-Positioning LoRA for Real Image Dehazing

• Link：https://arxiv.org/abs/2603.10872
• Code：https://github.com/YanZhang-zy/BiLaLoRA

UniRain: Unified Image Deraining with RAG-based Dataset Distillation and Multi-objective Reweighted Optimization

• Link：https://arxiv.org/abs/2603.03967
• Code：https://github.com/QianfengY/UniRain

具身智能（Embodied AI）

MSGNav: Unleashing the Power of Multi-modal 3D Scene Graph for Zero-Shot Embodied Navigation

• Link：https://arxiv.org/abs/2511.10376
• Code：https://github.com/ylwhxht/MSGNav

ForceVLA2: Unleashing Hybrid Force-Position Control with Force Awareness for Contact-Rich Manipulation

• Link：https://arxiv.org/abs/2603.15169
• Code：https://sites.google.com/view/force-vla2/home

RealVLG-R1: A Large-Scale Real-World Visual-Language Grounding Benchmark for Robotic Perception and Manipulation

• Link：https://arxiv.org/abs/2603.14880
• Code：https://github.com/lif314/RealVLG-R1

SaPaVe: Towards Active Perception and Manipulation in Vision-Language-Action Models for Robotics

• Link：https://arxiv.org/abs/2603.12193
• Code：https://lmzpai.github.io/SaPaVe

Context-Nav: Context-Driven Exploration and Viewpoint-Aware 3D Spatial Reasoning for Instance Navigation

• Link：https://arxiv.org/abs/2603.09506
• Code：

MergeVLA: Cross-Skill Model Merging Toward a Generalist Vision-Language-Action Agent

• Link：https://arxiv.org/abs/2511.18810
• Code：https://mergevla.github.io/

When Robots Obey the Patch: Universal Transferable Patch Attacks on Vision-Language-Action Models

• Link：https://arxiv.org/abs/2511.21192
• Code：

Context-Nav: Context-Driven Exploration and Viewpoint-Aware 3D Spatial Reasoning for Instance Navigation

• Link：https://arxiv.org/abs/2603.09506
• Code：https://github.com/AutoCompSysLab/ContextNav

Learning to See and Act: Task-Aware Virtual View Exploration for Robotic Manipulation

• Link：https://arxiv.org/abs/2508.05186
• Code：https://github.com/HCPLab-SYSU/TAVP.git

Cross-Domain Demo-to-Code via Neurosymbolic Counterfactual Reasoning

• Link：https://arxiv.org/abs/2603.18495
• Code：

Structural Action Transformer for 3D Dexterous Manipulation

• Link：https://arxiv.org/abs/2603.03960
• Code：

Action-Geometry Prediction with 3D Geometric Prior for Bimanual Manipulation

• Link：https://arxiv.org/abs/2602.23814
• Code：https://github.com/Chongyang-99/GAP.git

GeCo-SRT: Geometry-aware Continual Adaptation for Robotic Cross-Task Sim-to-Real Transfer

• Link：https://arxiv.org/abs/2602.20871
• Code：

Probing and Bridging Geometry-Interaction Cues for Affordance Reasoning in Vision Foundation Models

• Link：https://arxiv.org/abs/2602.20501
• Code：https://github.com/facebookresearch/affordance-probing

MSGNav: Unleashing the Power of Multi-modal 3D Scene Graph for Zero-Shot Embodied Navigation

• Link：https://arxiv.org/abs/2511.10376
• Code：

MindPower: Enabling Theory-of-Mind Reasoning in VLM-based Embodied Agents

• Link：https://arxiv.org/abs/2602.20412
• Code：

3DGS(Gaussian Splatting)

BrepGaussian: CAD reconstruction from Multi-View Images with Gaussian Splatting

• Link：https://arxiv.org/abs/2602.21105
• Code：

CrowdGaussian: Reconstructing High-Fidelity 3D Gaussians for Human Crowd from a Single Image

• Link：https://arxiv.org/abs/2603.17779
• Code：

ReLaGS: Relational Language Gaussian Splatting

• Link：https://arxiv.org/abs/2603.17605
• Code：https://dfki-av.github.io/ReLaGS/

Motion-Aware Animatable Gaussian Avatars Deblurring

• Link：https://arxiv.org/abs/2411.16758
• Code：https://github.com/MyNiuuu/MAD-Avatar

LTGS: Long-Term Gaussian Scene Chronology From Sparse View Updates

• Link：https://arxiv.org/abs/2510.09881
• Code：

EMGauss: Continuous Slice-to-3D Reconstruction via Dynamic Gaussian Modeling in Volume Electron Microscopy

• Link：https://arxiv.org/abs/2512.06684
• Code：https://raynehe.github.io/EMGauss/

PhyGaP: Physically-Grounded Gaussians with Polarization Cues

• Link：https://arxiv.org/abs/2603.14001
• Code：

Speeding Up the Learning of 3D Gaussians with Much Shorter Gaussian Lists

• Link：https://arxiv.org/abs/2603.09277
• Code：https://github.com/MachinePerceptionLab/ShorterSplatting

REALM: An MLLM-Agent Framework for Open World 3D Reasoning Segmentation and Editing on Gaussian Splatting

• Link：https://arxiv.org/abs/2510.16410
• Code：https://ChangyueShi.github.io/REALM

VarSplat: Uncertainty-aware 3D Gaussian Splatting for Robust RGB-D SLAM

• Link：https://arxiv.org/abs/2603.09673
• Code：

E2EGS: Event-to-Edge Gaussian Splatting for Pose-Free 3D Reconstruction

• Link：https://arxiv.org/abs/2603.14684
• Code：

ProgressiveAvatars: Progressive Animatable 3D Gaussian Avatars

• Link：https://arxiv.org/abs/2603.16447
• Code：https://ustc3dv.github.io/ProgressiveAvatars/

Let it Snow! Animating 3D Gaussian Scenes with Dynamic Weather Effects via Physics-Guided Score Distillation

• Link：https://arxiv.org/abs/2504.05296
• Code：https://galfiebelman.github.io/let-it-snow/

OnlineX: Unified Online 3D Reconstruction and Understanding with Active-to-Stable State Evolution

• Link：https://arxiv.org/abs/2603.02134
• Code：https://xiac20.github.io/OnlineX/

STAvatar: Soft Binding and Temporal Density Control for Monocular 3D Head Avatars Reconstruction

• Link：https://arxiv.org/abs/2511.19854
• Code：https://jiankuozhao.github.io/STAvatar/

Dropping Anchor and Spherical Harmonics for Sparse-view Gaussian Splatting

• Link：https://arxiv.org/abs/2602.20933
• Code：https://sk-fun.fun/DropAnSH-GS

RAP: Fast Feedforward Rendering-Free Attribute-Guided Primitive Importance Score Prediction for Efficient 3D Gaussian Splatting Processing

• Link：https://arxiv.org/abs/2602.19753
• Code：https://github.com/yyyykf/RAP

三维重建(3D Reconstruction)

GeodesicNVS: Probability Density Geodesic Flow Matching for Novel View Synthesis

• Link：https://arxiv.org/abs/2603.01010
• Code：

Talking Together: Synthesizing Co-Located 3D Conversations from Audio

• Link：https://arxiv.org/abs/2603.08674
• Code：

Parallelised Differentiable Straightest Geodesics for 3D Meshes

• Link：https://arxiv.org/abs/2603.15780
• Code：https://circle-group.github.io/research/DSG

FoV-Net: Rotation-Invariant CAD B-rep Learning via Field-of-View Ray Casting

• Link：https://arxiv.org/abs/2602.24084
• Code：

A2Z-10M+: Geometric Deep Learning with A-to-Z BRep Annotations for AI-Assisted CAD Modeling and Reverse Engineering

• Link：https://arxiv.org/abs/2603.12605
• Code：

Order Matters: 3D Shape Generation from Sequential VR Sketches

• Link：https://arxiv.org/abs/2512.04761
• Code：https://chenyizi086.github.io/VRSketch2Shape_website

CustomTex: High-fidelity Indoor Scene Texturing via Multi-Reference Customization

• Link：https://arxiv.org/abs/2603.19121
• Code：

PanoVGGT: Feed-Forward 3D Reconstruction from Panoramic Imagery

• Link：https://arxiv.org/abs/2603.17571
• Code：

SimRecon: SimReady Compositional Scene Reconstruction from Real Videos

• Link：https://arxiv.org/abs/2603.02133
• Code：https://xiac20.github.io/SimRecon/

MoRe: Motion-aware Feed-forward 4D Reconstruction Transformer

• Link：https://arxiv.org/abs/2603.05078
• Code：https://hellexf.github.io/MoRe/

RnG: A Unified Transformer for Complete 3D Modeling from Partial Observations

• Link：https://arxiv.org/abs/2603.01194
• Code：https://npucvr.github.io/RnG

tttLRM: Test-Time Training for Long Context and Autoregressive 3D Reconstruction

• Link：https://arxiv.org/abs/2602.20160
• Code：https://cwchenwang.github.io/tttLRM

FoV-Net: Rotation-Invariant CAD B-rep Learning via Field-of-View Ray Casting

• Link：https://arxiv.org/abs/2602.24084
• Code：

Global-Aware Edge Prioritization for Pose Graph Initialization

• Link：https://arxiv.org/abs/2602.21963
• Code：https://github.com/weitong8591/global_edge_prior

MoVieS: Motion-Aware 4D Dynamic View Synthesis in One Second

• Link：https://arxiv.org/abs/2507.10065
• Code：

tttLRM: Test-Time Training for Long Context and Autoregressive 3D Reconstruction

• Link：https://arxiv.org/abs/2602.20089
• Code：

模型剪枝(Model Pruning)

When Token Pruning is Worse than Random: Understanding Visual Token Information in VLLMs

• Link：https://arxiv.org/abs/2512.07580
• Code：https://github.com/YahongWang1/Information-Horizon

Prune2Drive: A Plug-and-Play Framework for Accelerating Vision-Language Models in Autonomous Driving

• Link：https://arxiv.org/abs/2508.13305
• Code：https://github.com/MinhaoXiong/Prune2Drive.git

Pluggable Pruning with Contiguous Layer Distillation for Diffusion Transformers

• Link：https://arxiv.org/abs/2511.16156
• Code：https://github.com/OPPO-Mente-Lab/Qwen-Image-Pruning

深度估计(Depth Estimation)

SpiderCam: Low-Power Snapshot Depth from Differential Defocus

• Link：https://arxiv.org/abs/2603.17910
• Code：

轨迹预测(Trajectory Prediction)

FoSS: Modeling Long Range Dependencies and Multimodal Uncertainty in Trajectory Prediction via Fourier State Space Integration

• Link：https://arxiv.org/abs/2603.01284
• Code：

Recover to Predict: Progressive Retrospective Learning for Variable-Length Trajectory Prediction

• Link：https://arxiv.org/abs/2603.10597
• Code：https://github.com/zhouhao94/PRF

Mamba / SSM

DA-Mamba: Learning Domain-Aware State Space Model for Global-Local Alignment in Domain Adaptive Object Detection

• Link：https://arxiv.org/abs/2603.18757
• Code：

Avatars

自动驾驶

AdaRadar: Rate Adaptive Spectral Compression for Radar-based Perception

• Link：https://arxiv.org/abs/2603.17979
• Code：

CausalVAD: De-confounding End-to-End Autonomous Driving via Causal Intervention

• Link：https://arxiv.org/abs/2603.18561
• Code：

SimScale: Learning to Drive via Real-World Simulation at Scale

• Link：https://arxiv.org/abs/2511.23369
• Code：https://github.com/OpenDriveLab/SimScale

VIRD: View-Invariant Representation through Dual-Axis Transformation for Cross-View Pose Estimation

• Link：https://arxiv.org/abs/2603.12918
• Code：

All Vehicles Can Lie: Efficient Adversarial Defense in Fully Untrusted-Vehicle Collaborative Perception

• Link：https://arxiv.org/abs/2603.08498
• Code：

HG-Lane: High-Fidelity Generation of Lane Scenes under Adverse Weather and Lighting Conditions without Re-annotation

• Link：https://arxiv.org/abs/2603.10128
• Code：https://github.com/zdc233/HG-Lane

KnowVal: A Knowledge-Augmented and Value-Guided Autonomous Driving System

• Link：https://arxiv.org/abs/2512.20299
• Code：

SABER: Spatially Consistent 3D Universal Adversarial Objects for BEV Detectors

• Link：https://arxiv.org/abs/2505.22499
• Code：

Learning to Drive is a Free Gift: Large-Scale Label-Free Autonomy Pretraining from Unposed In-The-Wild Videos

• Link：https://arxiv.org/abs/2602.22091
• Code：

DriverGaze360: OmniDirectional Driver Attention with Object-Level Guidance

• Link：https://arxiv.org/abs/2512.14266
• Code：https://dfki-av.github.io/drivergaze360

CoLC: Communication-Efficient Collaborative Perception with LiDAR Completion

• Link：https://arxiv.org/abs/2603.00682
• Code：https://github.com/CatOneTwo/CoLC

Learning to Drive is a Free Gift: Large-Scale Label-Free Autonomy Pretraining from Unposed In-The-Wild Videos

• Link：https://arxiv.org/abs/2602.22091
• Code：

Dr.Occ: Depth- and Region-Guided 3D Occupancy from Surround-View Cameras for Autonomous Driving

• Link：https://arxiv.org/abs/2603.01007
• Code：

LiREC-Net: A Target-Free and Learning-Based Network for LiDAR, RGB, and Event Calibration

• Link：https://arxiv.org/abs/2602.21754
• Code：

RAYNOVA: Scale-Temporal Autoregressive World Modeling in Ray Space

• Link：https://arxiv.org/abs/2602.20685
• Code：https://raynova-ai.github.io/

HorizonForge: Driving Scene Editing with Any Trajectories and Any Vehicles

• Link：https://arxiv.org/abs/2602.21333
• Code：https://horizonforge.github.io/

NoRD: A Data-Efficient Vision-Language-Action Model that Drives without Reasoning

• Link：https://arxiv.org/abs/2602.21172
• Code：

Perception Characteristics Distance: Measuring Stability and Robustness of Perception System in Dynamic Conditions under a Certain Decision Rule

• Link：https://arxiv.org/abs/2506.09217
• Code：https://github.com/datadrivenwheels/PCD_Python

SafeDrive: Fine-Grained Safety Reasoning for End-to-End Driving in a Sparse World

• Link：https://arxiv.org/abs/2602.18887
• Code：

Backbone

Hyperbolic Busemann Neural Networks

• Link：https://arxiv.org/abs/2602.18858
• Code：https://github.com/GitZH-Chen/HBNN

PFGNet: A Fully Convolutional Frequency-Guided Peripheral Gating Network for Efficient Spatiotemporal Predictive Learning

• Link：https://arxiv.org/abs/2602.20537
• Code：https://github.com/fhjdqaq/PFGNet

CLIP

FALCON: False-Negative Aware Learning of Contrastive Negatives in Vision-Language Alignment

• Link：https://arxiv.org/abs/2505.11192
• Code：

Reevaluating the Intra-Modal Misalignment Hypothesis in CLIP

• Link：https://arxiv.org/abs/2603.16100
• Code：

CHIPS: Efficient CLIP Adaptation via Curvature-aware Hybrid Influence-based Data Selection

• Link：https://arxiv.org/abs/2511.18519
• Code：

CLIPoint3D: Language-Grounded Few-Shot Unsupervised 3D Point Cloud Domain Adaptation

• Link：https://arxiv.org/abs/2602.20160
• Code：https://github.com/SarthakM320/CLIPoint3D

FluoCLIP: Stain-Aware Focus Quality Assessment in Fluorescence Microscopy

• Link：https://arxiv.org/abs/2602.23791
• Code：

MAE

Detecting AI-Generated Forgeries via Iterative Manifold Deviation Amplification

• Link：https://arxiv.org/abs/2602.18842
• Code：

SARMAE: Masked Autoencoder for SAR Representation Learning

• Link：https://arxiv.org/abs/2512.16635
• Code：https://github.com/MiliLab/SARMAE

OCR

What Is Wrong with Synthetic Data for Scene Text Recognition? A Strong Synthetic Engine with Diverse Simulations and Self-Evolution

• Link：https://arxiv.org/abs/2602.06450
• Code：https://github.com/YesianRohn/UnionST

TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering

• Link：https://arxiv.org/abs/2602.20903
• Code：https://github.com/CIawevy/TextPecker

Efficient Document Parsing via Parallel Token Prediction

• Link：https://arxiv.org/abs/2603.15206
• Code：

D2Dewarp: Dual Dimensions Geometric Representation Learning Based Document Image Dewarping

• Link：https://arxiv.org/abs/2507.08492
• Code：https://github.com/xiaomore/D2Dewarp

Occupancy

OneOcc: Semantic Occupancy Prediction for Legged Robots with a Single Panoramic Camera

• Link：https://arxiv.org/abs/2511.03571
• Code：https://github.com/MasterHow/OneOcc

Dr.Occ: Depth- and Region-Guided 3D Occupancy from Surround-View Cameras for Autonomous Driving

• Link：https://arxiv.org/abs/2603.01007
• Code：

NeRF

Spectral-Geometric Neural Fields for Pose-Free LiDAR View Synthesis

• Link：https://arxiv.org/abs/2603.12903
• Code：

Node-RF: Learning Generalized Continuous Space-Time Scene Dynamics with Neural ODE-based NeRFs

• Link：https://arxiv.org/abs/2603.12078
• Code：

Seeing through Light and Darkness: Sensor-Physics Grounded Deblurring HDR NeRF from Single-Exposure Images and Events

• Link：https://arxiv.org/abs/2601.15475
• Code：https://icvteam.github.io/See-NeRF.html

NERFIFY: A Multi-Agent Framework for Turning NeRF Papers into Code

• Link：https://arxiv.org/abs/2603.00805
• Code：

DETR

EW-DETR: Evolving World Object Detection via Incremental Low-Rank DEtection TRansformer

• Link：https://arxiv.org/abs/2602.20985
• Code：

GNN

Prompt

Towards Calibrating Prompt Tuning of Vision-Language Models

• Link：https://arxiv.org/abs/2602.19024
• Code：

PHAC: Promptable Human Amodal Completion

• Link：https://arxiv.org/abs/2603.14741
• Code：

FOZO: Forward-Only Zeroth-Order Prompt Optimization for Test-Time Adaptation

• Link：https://arxiv.org/abs/2603.04733
• Code：

FOZO: Forward-Only Zeroth-Order Prompt Optimization for Test-Time Adaptation

• Link：https://arxiv.org/abs/2603.04733
• Code：

大语言模型(LLM)

VecGlypher: Unified Vector Glyph Generation with Language Models

• Link：https://arxiv.org/abs/2602.21461
• Code：https://xk-huang.github.io/VecGlypher/

LaMoGen: Language to Motion Generation Through LLM-Guided Symbolic Inference

• Link：https://arxiv.org/abs/2603.11605
• Code：https://jjkislele.github.io/LaMoGen/

视觉语言模型(LLM)

Draft and Refine with Visual Experts

• Link：https://arxiv.org/abs/2511.11005
• Code：https://github.com/EavnJeong/Draft-and-Refine-with-Visual-Experts

Interpretable Debiasing of Vision-Language Models for Social Fairness

• Link：https://arxiv.org/abs/2602.24014
• Code：

Ego: Embedding-Guided Personalization of Vision-Language Models

• Link：https://arxiv.org/abs/2603.09771
• Code：

GTR-Turbo: Merged Checkpoint is Secretly a Free Teacher for Agentic VLM Training

• Link：https://arxiv.org/abs/2512.13043
• Code：

V-Attack: Targeting Disentangled Value Features for Controllable Adversarial Attacks on LVLMs

• Link：https://arxiv.org/abs/2511.20223
• Code：https://github.com/Summu77/V-Attack

It's Time to Get It Right: Improving Analog Clock Reading and Clock-Hand Spatial Reasoning in Vision-Language Models

• Link：https://arxiv.org/abs/2603.08011
• Code：

Mind the Way You Select Negative Texts: Pursuing the Distance Consistency in OOD Detection with VLMs

• Link：https://arxiv.org/abs/2603.02618
• Code：

AdaptVision: Efficient Vision-Language Models via Adaptive Visual Acquisition

• Link：https://arxiv.org/abs/2512.03794
• Code：https://github.com/AdaptVision/AdaptVision

DeAR: Fine-Grained VLM Adaptation by Decomposing Attention Head Roles

• Link：https://arxiv.org/abs/2603.01111
• Code：

Do Vision-Language Models Leak What They Learn? Adaptive Token-Weighted Model Inversion Attacks

• Link：https://arxiv.org/abs/2508.04097
• Code：https://ngoc-nguyen-0.github.io/SMI_AW/

Seeing Clearly, Reasoning Confidently: Plug-and-Play Remedies for Vision Language Model Blindness

• Link：https://arxiv.org/abs/2602.19615
• Code：

Quant Experts: Token-aware Adaptive Error Reconstruction with Mixture of Experts for Large Vision-Language Models Quantization

• Link：https://arxiv.org/abs/2602.24059
• Code：

多模态大语言模型(MLLM)

No Need For Real Anomaly: MLLM Empowered Zero-Shot Video Anomaly Detection

• Link：https://arxiv.org/abs/2602.19248
• Code：https://github.com/VitaminCreed/LAVIDA

See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles

• Link：https://arxiv.org/abs/2509.13615
• Code：https://github.com/ZrW00/StaR

Where MLLMs Attend and What They Rely On: Explaining Autoregressive Token Generation

• Link：https://arxiv.org/abs/2509.22496
• Code：

Fine-Grained Post-Training Quantization for Large Vision Language Models with Quantization-Aware Integrated Gradients

• Link：https://arxiv.org/abs/2603.17809
• Code：https://github.com/ucas-xiang/QIG

How to Take a Memorable Picture? Empowering Users with Actionable Feedback

• Link：https://arxiv.org/abs/2602.21877
• Code：https://laitifranz.github.io/MemCoach/

Rethinking MLLM Itself as a Segmenter with a Single Segmentation Token

• Link：https://arxiv.org/abs/2603.19026
• Code：https://github.com/ANDYZAQ/SELF1E

MRD: Multi-resolution Retrieval-Detection Fusion for High-Resolution Image Understanding

• Link：https://arxiv.org/abs/2512.02906
• Code：https://github.com/yf0412/MRD

OddGridBench: Exposing the Lack of Fine-Grained Visual Discrepancy Sensitivity in Multimodal Large Language Models

• Link：https://arxiv.org/abs/2603.09326
• Code：https://wwwtttjjj.github.io/OddGridBench/

FORCE: Transferable Visual Jailbreaking Attacks via Feature Over-Reliance CorrEction

• Link：https://arxiv.org/abs/2509.21029
• Code：

HIFICL: High-Fidelity In-Context Learning for Multimodal Tasks

• Link：https://arxiv.org/abs/2603.12760
• Code：https://github.com/bbbandari/HiFICL

Parallel In-context Learning for Large Vision Language Models

• Link：https://arxiv.org/abs/2603.16092
• Code：https://github.com/yshinya6/parallel-icl

Multi-Crit: Benchmarking Multimodal Judges on Pluralistic Criteria-Following

• Link：https://arxiv.org/abs/2511.21662
• Code：

SPARROW: Learning Spatial Precision and Temporal Referential Consistency in Pixel-Grounded Video MLLMs

• Link：https://arxiv.org/abs/2603.12382
• Code：https://risys-lab.github.io/SPARROW；https://github.com/RISys-Lab/SPARROW

Tokenization Allows Multimodal Large Language Models to Understand, Generate and Edit Architectural Floor Plans

• Link：https://arxiv.org/abs/2603.11640
• Code：

WalkGPT: Grounded Vision-Language Conversation with Depth-Aware Segmentation for Pedestrian Navigation

• Link：https://arxiv.org/abs/2603.10703
• Code：https://sites.google.com/view/walkgpt-26/home

The Coherence Trap: When MLLM-Crafted Narratives Exploit Manipulated Visual Contexts

• Link：https://arxiv.org/abs/2505.17476
• Code：

KVSmooth: Mitigating Hallucination in Multi-modal Large Language Models through Key-Value Smoothing

• Link：https://arxiv.org/abs/2602.04268
• Code：

IMAIA: Interactive Maps AI Assistant for Travel Planning and Geo-Spatial Intelligence

• Link：https://arxiv.org/abs/2507.06993
• Code：

LLaVAShield: Safeguarding Multimodal Multi-Turn Dialogues in Vision-Language Models

• Link：https://arxiv.org/abs/2509.25896
• Code：

Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memory

• Link：https://arxiv.org/abs/2603.15800
• Code：https://echosafe-mllm.github.io

Locate-then-Sparsify: Attribution Guided Sparse Strategy for Visual Hallucination Mitigation

• Link：https://arxiv.org/abs/2603.16284
• Code：

GUI-CEval: A Hierarchical and Comprehensive Chinese Benchmark for Mobile GUI Agents

• Link：https://arxiv.org/abs/2603.15039
• Code：

Rationale-Enhanced Decoding for Multi-modal Chain-of-Thought

• Link：https://arxiv.org/abs/2507.07685
• Code：https://github.com/yshinya6/red/

ViRC: Enhancing Visual Interleaved Mathematical CoT with Reason Chunking

• Link：https://arxiv.org/abs/2512.14654
• Code：https://github.com/Leon-LihongWang/ViRC

See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles

• Link：https://arxiv.org/abs/2509.13615
• Code：https://github.com/ZrW00/StaR

Evolutionary Multimodal Reasoning via Hierarchical Semantic Representation for Intent Recognition

• Link：https://arxiv.org/abs/2603.03827
• Code：https://github.com/thuiar/HIER

Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation

• Link：https://arxiv.org/abs/2511.22235
• Code：https://github.com/hehehahi4/CES

Graph2Eval: Automatic Multimodal Task Generation for Agents via Knowledge Graphs

• Link：https://arxiv.org/abs/2510.00507
• Code：

MASQuant: Modality-Aware Smoothing Quantization for Multimodal Large Language Models

• Link：https://arxiv.org/abs/2603.04800
• Code：https://github.com/alibaba/EfficientAI

EMO-R3: Reflective Reinforcement Learning for Emotional Reasoning in Multimodal Large Language Models

• Link：https://arxiv.org/abs/2602.23802
• Code：

ReMoRa: Multimodal Large Language Model based on Refined Motion Representation for Long-Video Understanding

• Link：https://arxiv.org/abs/2602.16412
• Code：

Venus: Benchmarking and Empowering Multimodal Large Language Models for Aesthetic Guidance and Cropping

• Link：https://arxiv.org/abs/2602.23980
• Code：https://github.com/PKU-ICST-MIPL/Venus_CVPR2026

WeaveTime: Stream from Earlier Frames into Emergent Memory in VideoLLMs

• Link：https://arxiv.org/abs/2602.22142
• Code：https://zhangyl4.github.io/publications/weavetime/

MoDES: Accelerating Mixture-of-Experts Multimodal Large Language Models via Dynamic Expert Skipping

• Link：https://arxiv.org/abs/2511.15690
• Code：https://github.com/ModelTC/MoDES

Echoes of Ownership: Adversarial-Guided Dual Injection for Copyright Protection in MLLMs

• Link：https://arxiv.org/abs/2602.18845
• Code：

多模态

ConsistCompose: Unified Multimodal Layout Control for Image Composition

• Link：https://arxiv.org/abs/2511.18333
• Code：

Mixture of States: Routing Token-Level Dynamics for Multimodal Generation

• Link：https://arxiv.org/abs/2511.12207
• Code：https://haozheliu-st.github.io/mos-homepage/

Intrinsic Concept Extraction Based on Compositional Interpretability

• Link：https://arxiv.org/abs/2603.11795
• Code：

ParTY: Part-Guidance for Expressive Text-to-Motion Synthesis

• Link：https://arxiv.org/abs/2603.09611
• Code：https://github.com/VisualScienceLab-KHU/ParTY

x²-Fusion: Cross-Modality and Cross-Dimension Flow Estimation in Event Edge Space

• Link：https://arxiv.org/abs/2603.16671
• Code：

EI: Early Intervention for Multimodal Imaging based Disease Recognition

• Link：https://arxiv.org/abs/2603.17514
• Code：

Decoupling Stability and Plasticity for Multi-Modal Test-Time Adaptation

• Link：https://arxiv.org/abs/2603.00574
• Code：

Linking Modality Isolation in Heterogeneous Collaborative Perception

• Link：https://arxiv.org/abs/2603.00609
• Code：https://github.com/cxliu0314/CodeAlign

UniMMAD: Unified Multi-Modal and Multi-Class Anomaly Detection via MoE-Driven Feature Decompression

• Link：https://arxiv.org/abs/2509.25934
• Code：https://github.com/yuanzhao-CVLAB/UniMMAD

Beyond Global Similarity: Towards Fine-Grained, Multi-Condition Multimodal Retrieval

• Link：https://arxiv.org/abs/2603.01082
• Code：https://github.com/EIT-NLP/MCMR

Adaptive Confidence Regularization for Multimodal Failure Detection

• Link：https://arxiv.org/abs/2603.02200
• Code：https://github.com/mona4399/ACR

Cross-modal Identity Mapping: Minimizing Information Loss in Modality Conversion via Reinforcement Learning

• Link：https://arxiv.org/abs/2603.01696
• Code：

VideoFusion: A Spatio-Temporal Collaborative Network for Multi-modal Video Fusion

• Link：https://arxiv.org/abs/2503.23359
• Code：https://github.com/Linfeng-Tang/VideoFusion

U-Mind: A Unified Framework for Real-Time Multimodal Interaction with Audiovisual Generation

• Link：https://arxiv.org/abs/2602.23739
• Code：

MultiModalPFN: Extending Prior-Data Fitted Networks for Multimodal Tabular Learning

• Link：https://arxiv.org/abs/2602.20223
• Code：https://github.com/too-z/MultiModalPFN

Multi-Modal Representation Learning via Semi-Supervised Rate Reduction for Generalized Category Discovery

• Link：https://arxiv.org/abs/2602.19910
• Code：

CLCR: Cross-Level Semantic Collaborative Representation for Multimodal Learning

• Link：https://arxiv.org/abs/2602.19605
• Code：

Tri-Subspaces Disentanglement for Multimodal Sentiment Analysis

• Link：https://arxiv.org/abs/2602.19585
• Code：

CaReFlow: Cyclic Adaptive Rectified Flow for Multimodal Fusion

• Link：https://arxiv.org/abs/2602.19140
• Code：

NAS

视觉问答(Visual Question Answering)

Step-CoT: Stepwise Visual Chain-of-Thought for Medical Visual Question Answering

• Link：https://arxiv.org/abs/2603.13878
• Code：github.com/hahaha111111/Step-CoT

Do You See What I Am Pointing At? Gesture-Based Egocentric Video Question Answering

• Link：https://arxiv.org/abs/2603.12533
• Code：https://yuuraa.github.io/papers/choi2026egovqa

SpatiaLQA: A Benchmark for Evaluating Spatial Logical Reasoning in Vision-Language Models

• Link：https://arxiv.org/abs/2602.20901
• Code：https://github.com/xieyc99/SpatiaLQA

强化学习(Reinforcement Learning)

Dual-Agent Reinforcement Learning for Adaptive and Cost-Aware Visual-Inertial Odometry

• Link：https://arxiv.org/abs/2511.21083
• Code：

Learning to Assist: Physics-Grounded Human-Human Control via Multi-Agent Reinforcement Learning

• Link：https://arxiv.org/abs/2603.11346
• Code：https://yutoshibata07.github.io/AssistMimic-projectpage/

RL-ScanIQA: Reinforcement-Learned Scanpaths for Blind 360°Image Quality Assessment

• Link：https://arxiv.org/abs/2603.14297
• Code：https://github.com/wangyuji1/RLScanIQA.git

Specificity-aware reinforcement learning for fine-grained open-world classification

• Link：https://arxiv.org/abs/2603.03197
• Code：https://github.com/s-angheben/SpeciaRL

From Pairs to Sequences: Track-Aware Policy Gradients for Keypoint Detection

• Link：https://arxiv.org/abs/2602.20630
• Code：

OraPO: Oracle-educated Reinforcement Learning for Data-efficient and Factual Radiology Report Generation

• Link：https://arxiv.org/abs/2509.18600
• Code：

ReID(重识别)

长尾分布(Long-Tail)

Hier-COS: Making Deep Features Hierarchy-aware via Composition of Orthogonal Subspaces

• Link：https://arxiv.org/abs/2602.20068
• Code：

Meta-Learning Hyperparameters for Parameter Efficient Fine-Tuning

• Link：https://arxiv.org/abs/2603.01759
• Code：https://github.com/doem97/metalora

视频压缩(Video Compression)

扩散模型(Diffusion Models)

Prototype-Guided Concept Erasure in Diffusion Models

• Link：https://arxiv.org/abs/2603.08271
• Code：

Pixel Motion Diffusion is What We Need for Robot Control

• Link：https://arxiv.org/abs/2509.22652
• Code：https://eronguyen.github.io/DAWN

CFG-Ctrl: Control-Based Classifier-Free Diffusion Guidance

• Link：https://arxiv.org/abs/2603.03281
• Code：https://hanyang-21.github.io/CFG-Ctrl

TAUE: Training-free Noise Transplant and Cultivation Diffusion Model

• Link：https://arxiv.org/abs/2511.02580
• Code：https://iyatomilab.github.io/TAUE

3M-TI: High-Quality Mobile Thermal Imaging via Calibration-free Multi-Camera Cross-Modal Diffusion

• Link：https://arxiv.org/abs/2511.19117
• Code：https://github.com/work-submit/3MTI

DiFlowDubber: Discrete Flow Matching for Automated Video Dubbing via Cross-Modal Alignment and Synchronization

• Link：https://arxiv.org/abs/2603.14267
• Code：

CoD: A Diffusion Foundation Model for Image Compression

• Link：https://arxiv.org/abs/2511.18706
• Code：https://github.com/microsoft/GenCodec/tree/main/CoD

SpiralDiff: Spiral Diffusion with LoRA for RGB-to-RAW Conversion Across Cameras

• Link：https://arxiv.org/abs/2603.14885
• Code：https://github.com/Chuancy-TJU/SpiralDiff

All-in-One Slider for Attribute Manipulation in Diffusion Models

• Link：https://arxiv.org/abs/2508.19195
• Code：

LESA: Learnable Stage-Aware Predictors for Diffusion Model Acceleration

• Link：https://arxiv.org/abs/2602.20497
• Code：

When LoRA Betrays: Backdooring Text-to-Image Models by Masquerading as Benign Adapters

• Link：https://arxiv.org/abs/2602.21977
• Code：

SODA: Sensitivity-Oriented Dynamic Acceleration for Diffusion Transformer

• Link：https://arxiv.org/abs/2603.07057
• Code：https://github.com/leaves162/SODA

Guiding Diffusion Models with Semantically Degraded Conditions

• Link：https://arxiv.org/abs/2603.10780
• Code：https://github.com/Ming-321/Classifier-Degradation-Guidance

COT-FM: Cluster-wise Optimal Transport Flow Matching

• Link：https://arxiv.org/abs/2603.13395
• Code：

Taming Preference Mode Collapse via Directional Decoupling Alignment in Diffusion Reinforcement Learning

• Link：https://arxiv.org/abs/2512.24146
• Code：

Making Training-Free Diffusion Segmentors Scale with the Generative Power

• Link：https://arxiv.org/abs/2603.06178
• Code：https://github.com/Darkbblue/goca

Uni-DAD: Unified Distillation and Adaptation of Diffusion Models for Few-step Few-shot Image Generation

• Link：https://arxiv.org/abs/2511.18281
• Code：https://github.com/yaramohamadi/uni-DAD

Refining Few-Step Text-to-Multiview Diffusion via Reinforcement Learning

• Link：https://arxiv.org/abs/2505.20107
• Code：https://github.com/ZiyiZhang27/MVC-ZigAL

Face2Scene: Using Facial Degradation as an Oracle for Diffusion-Based Scene Restoration

• Link：https://arxiv.org/abs/2603.16570
• Code：

Cubic Discrete Diffusion: Discrete Visual Generation on High-Dimensional Representation Tokens

• Link：https://arxiv.org/abs/2603.19232
• Code：https://github.com/YuqingWang1029/CubiD

Adaptive Auxiliary Prompt Blending for Target-Faithful Diffusion Generation

• Link：https://arxiv.org/abs/2603.19158
• Code：

ADAPT: Attention Driven Adaptive Prompt Scheduling and InTerpolating Orthogonal Complements for Rare Concepts Generation

• Link：https://arxiv.org/abs/2603.19157
• Code：

All-in-One Slider for Attribute Manipulation in Diffusion Models

• Link：https://arxiv.org/abs/2508.19195
• Code：

TINA: Text-Free Inversion Attack for Unlearned Text-to-Image Diffusion Models

• Link：https://arxiv.org/abs/2603.17828
• Code：

Elucidating the Design Space of Arbitrary-Noise-Based Diffusion Models

• Link：https://arxiv.org/abs/2507.18534
• Code：https://github.com/PerceptionComputingLab/EDA

TAP: A Token-Adaptive Predictor Framework for Training-Free Diffusion Acceleration

• Link：https://arxiv.org/abs/2603.03792
• Code：

CFG-Ctrl: Control-Based Classifier-Free Diffusion Guidance

• Link：https://arxiv.org/abs/2603.03281
• Code：https://hanyang-21.github.io/CFG-Ctrl

ConceptPrism: Concept Disentanglement in Personalized Diffusion Models via Residual Token Optimization

• Link：https://arxiv.org/abs/2602.19575
• Code：

SeaCache: Spectral-Evolution-Aware Cache for Accelerating Diffusion Models

• Link：https://arxiv.org/abs/2602.18993
• Code：

Vision Transformer

Make it SING: Analyzing Semantic Invariants in Classifiers

• Link：https://arxiv.org/abs/2603.14610
• Code：

Revisiting Model Stitching In the Foundation Model Era

• Link：https://arxiv.org/abs/2603.12433
• Code：

BinaryAttention: One-Bit QK-Attention for Vision and Diffusion Transformers

• Link：https://arxiv.org/abs/2603.09582
• Code：https://github.com/EdwardChasel/BinaryAttention

MuViT: Multi-Resolution Vision Transformers for Learning Across Scales in Microscopy

• Link：https://arxiv.org/abs/2602.24222
• Code：

全景分割(Panoptic Segmentation)

Seeing Beyond: Extrapolative Domain Adaptive Panoramic Segmentation

• Link：https://arxiv.org/abs/2603.15475
• Code：https://github.com/zyfone/EDA-PSeg

视觉和语言(Vision-Language)

VL-RouterBench: A Benchmark for Vision-Language Model Routing

• Link：https://arxiv.org/abs/2512.23562
• Code：

CrossHOI-Bench: A Unified Benchmark for HOI Evaluation across Vision-Language Models and HOI-Specific Methods

• Link：https://arxiv.org/abs/2508.18753
• Code：

More than the Sum: Panorama-Language Models for Adverse Omni-Scenes

• Link：https://arxiv.org/abs/2603.09573
• Code：https://github.com/InSAI-Lab/PanoVQA

A Unified Benchmark for HOI Evaluation across Vision-Language Models and HOI-Specific Methods

• Link：https://arxiv.org/abs/2508.18753
• Code：

Probing and Bridging Geometry-Interaction Cues for Affordance Reasoning in Vision Foundation Models

• Link：https://arxiv.org/abs/2602.20501
• Code：

ProFocus: Proactive Perception and Focused Reasoning in Vision-and-Language Navigation

• Link：https://arxiv.org/abs/2603.05530
• Code：

AVION: Aerial Vision-Language Instruction from Offline Teacher to Prompt-Tuned Network

• Link：https://arxiv.org/abs/2603.12659
• Code：

HoneyBee: Data Recipes for Vision-Language Reasoners

• Link：https://arxiv.org/abs/2510.12225
• Code：https://huggingface.co/datasets/facebook/HoneyBee

HATS: Hardness-Aware Trajectory Synthesis for GUI Agents

• Link：https://arxiv.org/abs/2603.12138
• Code：

StructXLIP: Enhancing Vision-language Models with Multimodal Structural Cues

• Link：https://arxiv.org/abs/2602.20089
• Code：https://github.com/intelligolabs/StructXLIP

StructXLIP: Enhancing Vision-language Models with Multimodal Structural Cues

• Link：https://arxiv.org/abs/2503.07853
• Code：https://github.com/intelligolabs/StructXLIP

目标检测(Object Detection)

Prompt-Free Universal Region Proposal Network

• Link：https://arxiv.org/abs/2603.17554
• Code：https://github.com/tangqh03/PF-RPN

Does YOLO Really Need to See Every Training Image in Every Epoch?

• Link：https://arxiv.org/abs/2603.17684
• Code：

Fourier Angle Alignment for Oriented Object Detection in Remote Sensing

• Link：https://arxiv.org/abs/2602.23790
• Code：https://github.com/gcy0423/Fourier-Angle-Alignment

Fourier Angle Alignment for Oriented Object Detection in Remote Sensing

• Link：https://arxiv.org/abs/2602.23790
• Code：https://github.com/gcy0423/Fourier-Angle-Alignment

Foundation Model Priors Enhance Object Focus in Feature Space for Source-Free Object Detection

• Link：https://arxiv.org/abs/2512.17514
• Code：

数据增强(Data Augmentation)

Fixed Anchors Are Not Enough: Dynamic Retrieval and Persistent Homology for Dataset Distillation

• Link：https://arxiv.org/abs/2602.24144
• Code：

异常检测(Anomaly Detection)

MoECLIP: Patch-Specialized Experts for Zero-shot Anomaly Detection

• Link：https://arxiv.org/abs/2603.03101
• Code：https://github.com/CoCoRessa/MoECLIP

EReCu: Pseudo-label Evolution Fusion and Refinement with Multi-Cue Learning for Unsupervised Camouflage Detection

• Link：https://arxiv.org/abs/2603.11521
• Code：

RC-NF: Robot-Conditioned Normalizing Flow for Real-Time Anomaly Detection in Robotic Manipulation

• Link：https://arxiv.org/abs/2603.11106
• Code：

Mind the Way You Select Negative Texts: Pursuing the Distance Consistency in OOD Detection with VLMs

• Link：https://arxiv.org/abs/2603.02618
• Code：

No Need For Real Anomaly: MLLM Empowered Zero-Shot Video Anomaly Detection

• Link：https://arxiv.org/abs/2602.19248
• Code：https://github.com/VitaminCreed/LAVIDA

UniMMAD: Unified Multi-Modal and Multi-Class Anomaly Detection via MoE-Driven Feature Decompression

• Link：https://arxiv.org/abs/2509.25934
• Code：https://github.com/yuanzhao-CVLAB/UniMMAD

Training-free Detection of Generated Videos via Spatial-Temporal Likelihoods

• Link：https://arxiv.org/abs/2603.15026
• Code：https://omerbenhayun.github.io/stall-video

VisualAD: Language-Free Zero-Shot Anomaly Detection via Vision Transformer

• Link：https://arxiv.org/abs/2603.07952
• Code：https://github.com/7HHHHH/VisualAD

Weakly Supervised Video Anomaly Detection with Anomaly-Connected Components and Intention Reasoning

• Link：https://arxiv.org/abs/2603.00550
• Code：

Diversity over Uniformity: Rethinking Representation in Generated Image Detection

• Link：https://arxiv.org/abs/2603.00717
• Code：https://github.com/Yanmou-Hui/DoU

The Invisible Gorilla Effect in Out-of-distribution Detection

• Link：https://arxiv.org/abs/2602.19944
• Code：https://github.com/HarryAnthony/Invisible_Gorilla_Effect

GS-CLIP: Zero-shot 3D Anomaly Detection by Geometry-Aware Prompt and Synergistic View Representation Learning

• Link：https://arxiv.org/abs/2602.19206
• Code：https://github.com/zhushengxinyue/GS-CLIP

No Need For Real Anomaly: MLLM Empowered Zero-Shot Video Anomaly Detection

• Link：https://arxiv.org/abs/2602.19248
• Code：https://github.com/VitaminCreed/LAVIDA

SimLBR: Learning to Detect Fake Images by Learning to Detect Real Images

• Link：https://arxiv.org/pdf/2602.20412
• Code：

目标跟踪(Object Tracking)

UTPTrack: Towards Simple and Unified Token Pruning for Visual Tracking

• Link：https://arxiv.org/abs/2602.23734
• Code：https://github.com/EIT-NLP/UTPTrack

Rethinking Two-Stage Referring-by-Tracking in Referring Multi-Object Tracking

• Link：https://arxiv.org/abs/2503.07516
• Code：https://github.com/buptLwz/FlexHook

Occlusion-Aware SORT: Observing Occlusion for Robust Multi-Object Tracking

• Link：https://arxiv.org/abs/2603.06034
• Code：

Changes in Real Time: Online Scene Change Detection with Multi-View Fusion

• Link：https://arxiv.org/abs/2511.12370
• Code：https://chumsy0725.github.io/O-SCD/

Learning Cross-View Object Correspondence via Cycle-Consistent Mask Prediction

• Link：https://arxiv.org/abs/2602.18996
• Code：https://github.com/shannany0606/CCMP

语义分割(Semantic Segmentation)

ACPV-Net: All-Class Polygonal Vectorization for Seamless Vector Map Generation from Aerial Imagery

• Link：https://arxiv.org/abs/2603.16671
• Code：https://github.com/HeinzJiao/ACPV-Net

Towards High-Quality Image Segmentation: Improving Topology Accuracy by Penalizing Neighbor Pixels

• Link：https://arxiv.org/abs/2603.18671
• Code：https://jmlipman.github.io/SCNP-SameClassNeighborPenalization

MixerCSeg: An Efficient Mixer Architecture for Crack Segmentation via Decoupled Mamba Attention

• Link：https://arxiv.org/abs/2603.01361
• Code：https://github.com/spiderforest/MixerCSeg

Shape-of-You: Fused Gromov-Wasserstein Optimal Transport for Semantic Correspondence in-the-Wild

• Link：https://arxiv.org/abs/2603.11618
• Code：

Discriminative Perception via Anchored Description for Reasoning Segmentation

• Link：https://arxiv.org/abs/2603.04002
• Code：https://github.com/mrazhou/DPAD

实例分割(Instance Segmentation)

少样本学习(Few-Shot Learning)

Remedying Target-Domain Astigmatism for Cross-Domain Few-Shot Object Detection

• Link：https://arxiv.org/abs/2603.18541
• Code：

MAGIC: Few-Shot Mask-Guided Anomaly Inpainting with Prompt Perturbation, Spatially Adaptive Guidance, and Context Awareness

• Link：https://arxiv.org/abs/2507.02314
• Code：https://github.com/SpatialAILab/MAGIC-Anomaly-generation

SCOPE: Scene-Contextualized Incremental Few-Shot 3D Segmentation

• Link：https://arxiv.org/abs/2603.06572
• Code：https://github.com/Surrey-UP-Lab/SCOPE

MUSE: Harnessing Precise and Diverse Semantics for Few-Shot Whole Slide Image Classification

• Link：https://arxiv.org/abs/2602.20873
• Code：https://github.com/JiahaoXu-god/CVPR2026_MUSE

Learning Multi-Modal Prototypes for Cross-Domain Few-Shot Object Detection

• Link：https://arxiv.org/abs/2602.18811
• Code：

生物医学

医学图像(Medical Image)

Sparse Task Vector Mixup with Hypernetworks for Efficient Knowledge Transfer in Whole-Slide Image Prognosis

• Link：https://arxiv.org/abs/2603.10526
• Code：https://github.com/liupei101/STEPH

CARE: A Molecular-Guided Foundation Model with Adaptive Region Modeling for Whole Slide Image Analysis

• Link：https://arxiv.org/abs/2602.21637
• Code：

CARE: A Molecular-Guided Foundation Model with Adaptive Region Modeling for Whole Slide Image Analysis

• Link：https://arxiv.org/abs/2602.21637
• Code：

Virtual Full-stack Scanning of Brain MRI via Imputing Any Quantised Code

• Link：https://arxiv.org/abs/2501.18328
• Code：https://github.com/ycwu1997/CodeBrain

LUMINA: A Multi-Vendor Mammography Benchmark with Energy Harmonization Protocol

• Link：https://arxiv.org/abs/2603.14644
• Code：

Towards Efficient Medical Reasoning with Minimal Fine-Tuning Data

• Link：https://arxiv.org/abs/2508.01450
• Code：https://github.com/mihara-bot/DIQ

Every Error has Its Magnitude: Asymmetric Mistake Severity Training for Multiclass Multiple Instance Learning

• Link：https://arxiv.org/abs/2603.13682
• Code：

X-WIN: Building Chest Radiograph World Model via Predictive Sensing

• Link：https://arxiv.org/abs/2511.14918
• Code：

Similarity-as-Evidence: Calibrating Overconfident VLMs for Interpretable and Label-Efficient Medical Active Learning

• Link：https://arxiv.org/abs/2602.18867
• Code：

LUMINA: A Multi-Vendor Mammography Benchmark with Energy Harmonization Protocol

• Link：https://arxiv.org/abs/2603.14644
• Code：

Cell-Type Prototype-Informed Neural Network for Gene Expression Estimation from Pathology Images

• Link：https://arxiv.org/abs/2603.18461
• Code：https://github.com/naivete5656/CPNN

Benchmarking Endoscopic Surgical Image Restoration and Beyond

• Link：https://arxiv.org/abs/2505.19161
• Code：

MRI Contrast Enhancement Kinetics World Model

• Link：https://arxiv.org/abs/2602.19285
• Code：https://github.com/DD0922/MRI-Contrast-Enhancement-Kinetics-World-Model

Similarity-as-Evidence: Calibrating Overconfident VLMs for Interpretable and Label-Efficient Medical Active Learning

• Link：https://arxiv.org/abs/2602.18867
• Code：

Solving a Nonlinear Blind Inverse Problem for Tagged MRI with Physics and Deep Generative Priors

• Link：https://arxiv.org/abs/2603.00882
• Code：

Act Like a Pathologist: Tissue-Aware Whole Slide Image Reasoning

• Link：https://arxiv.org/abs/2603.00667
• Code：

医学图像分割(Medical Image Segmentation)

Tell2Adapt: A Unified Framework for Source Free Unsupervised Domain Adaptation via Vision Foundation Model

• Link：https://arxiv.org/abs/2603.05012
• Code：https://github.com/derekshiii/Tell2Adapt

SPEGC: Continual Test-Time Adaptation via Semantic-Prompt-Enhanced Graph Clustering for Medical Image Segmentation

• Link：https://arxiv.org/abs/2603.11492
• Code：https://github.com/Jwei-Z/SPEGC-for-MIS

视频目标分割(Video Object Segmentation)

MatAnyone 2: Scaling Video Matting via a Learned Quality Evaluator

• Link：https://arxiv.org/abs/2512.11782
• Code：https://pq-yang.github.io/projects/MatAnyone2/

Synergistic Bleeding Region and Point Detection in Laparoscopic Surgical Videos

• Link：https://arxiv.org/abs/2503.22174
• Code：

行为检测(Action Detection)

SAVA-X: Ego-to-Exo Imitation Error Detection via Scene-Adaptive View Alignment and Bidirectional Cross View Fusion

• Link：https://arxiv.org/abs/2603.12764
• Code：https://github.com/jack1ee/SAVAX

人脸识别(Face Recognition)

RecoverMark: Robust Watermarking for Localization and Recovery of Manipulated Faces

• Link：https://arxiv.org/abs/2602.20618
• Code：

IDperturb: Enhancing Variation in Synthetic Face Generation via Angular Perturbation

• Link：https://arxiv.org/abs/2602.18831
• Code：

3D点云(3D-Point-Cloud)

Learning Coordinate-based Convolutional Kernels for Continuous SE(3) Equivariant and Efficient Point Cloud Analysis

• Link：https://arxiv.org/abs/2603.17538
• Code：

Points-to-3D: Structure-Aware 3D Generation with Point Cloud Priors

• Link：https://arxiv.org/abs/2603.18782
• Code：

Universal 3D Shape Matching via Coarse-to-Fine Language Guidance

• Link：https://arxiv.org/abs/2511.23055
• Code：

QD-PCQA: Quality-Aware Domain Adaptation for Point Cloud Quality Assessment

• Link：https://arxiv.org/abs/2603.03726
• Code：

QD-PCQA: Quality-Aware Domain Adaptation for Point Cloud Quality Assessment

• Link：https://arxiv.org/abs/2603.03726
• Code：https://github.com/huhu-code/QD-PCQA

自监督学习(Self-supervised Learning)

BEV-SLD: Self-Supervised Scene Landmark Detection for Global Localization with LiDAR Bird's-Eye View Images

• Link：https://arxiv.org/abs/2603.17159
• Code：

Towards Stable Self-Supervised Object Representations in Unconstrained Egocentric Video

• Link：https://arxiv.org/abs/2603.13912
• Code：

Bootstrap Dynamic-Aware 3D Visual Representation for Scalable Robot Learning

• Link：https://arxiv.org/abs/2512.00074
• Code：https://kolakivy.github.io/AFRO/

生物工程(bioengineering)

Multimodal Protein Language Models for Enzyme Kinetic Parameters: From Substrate Recognition to Conformational Adaptation

• Link：https://arxiv.org/abs/2603.12845
• Code：

联邦学习(Federated Learning)

Federated Active Learning Under Extreme Non-IID and Global Class Imbalance

• Link：https://arxiv.org/abs/2603.10341
• Code：https://github.com/chenchenzong/FairFAL

Domain-Skewed Federated Learning with Feature Decoupling and Calibration

• Link：https://arxiv.org/abs/2603.14238
• Code：https://github.com/mala-lab/F2DC

Fed-ADE: Adaptive Learning Rate for Federated Post-adaptation under Distribution Shift

• Link：https://arxiv.org/abs/2603.01040
• Code：

HiLoRA: Hierarchical Low-Rank Adaptation for Personalized Federated Learning

• Link：https://arxiv.org/abs/2603.02785
• Code：

FedVG: Gradient-Guided Aggregation for Enhanced Federated Learning

• Link：https://arxiv.org/abs/2602.21399
• Code：https://github.com/alinadevkota/FedVG

FedAFD: Multimodal Federated Learning via Adversarial Fusion and Distillation

• Link：https://arxiv.org/abs/2603.04890
• Code：

增量学习(Incremental Learning)

3D目标检测(3D Object Detection)

Look Before You Fuse: 2D-Guided Cross-Modal Alignment for Robust 3D Detection

• Link：https://arxiv.org/abs/2507.16861
• Code：

VirPro: Visual-referred Probabilistic Prompt Learning for Weakly-Supervised Monocular 3D Detection

• Link：https://arxiv.org/abs/2603.17470
• Code：

CoIn3D: Revisiting Configuration-Invariant Multi-Camera 3D Object Detection

• Link：https://arxiv.org/abs/2603.05042
• Code：

R4Det: 4D Radar-Camera Fusion for High-Performance 3D Object Detection

• Link：https://arxiv.org/abs/2603.11566
• Code：

SPAN: Spatial-Projection Alignment for Monocular 3D Object Detection

• Link：https://arxiv.org/abs/2511.06702
• Code：

Learning Mutual View Information Graph for Adaptive Adversarial Collaborative Perception

• Link：https://arxiv.org/abs/2602.19596
• Code：https://github.com/yihangtao/MVIG.git

SABER: Spatially Consistent 3D Universal Adversarial Objects for BEV Detectors

• Link：https://arxiv.org/abs/2505.22499
• Code：

VGGT-Det: Mining VGGT Internal Priors for Sensor-Geometry-Free Multi-View Indoor 3D Object Detection

• Link：https://arxiv.org/abs/2603.00912
• Code：https://github.com/yangcaoai/VGGT-Det-CVPR2026

3D语义分割(3D Semantic Segmentation)

图像编辑(Image Editing)

Precise Object and Effect Removal with Adaptive Target-Aware Attention

• Link：https://arxiv.org/abs/2505.22636
• Code：https://zjx0101.github.io/projects/ObjectClear/

CARE-Edit: Condition-Aware Routing of Experts for Contextual Image Editing

• Link：https://arxiv.org/abs/2603.08589
• Code：https://care-edit.github.io/

Rel-Zero: Harnessing Patch-Pair Invariance for Robust Zero-Watermarking Against AI Editing

• Link：https://arxiv.org/abs/2603.17531
• Code：

Cycle-Consistent Tuning for Layered Image Decomposition

• Link：https://arxiv.org/abs/2602.20989
• Code：

BeautyGRPO: Aesthetic Alignment for Face Retouching via Dynamic Path Guidance and Fine-Grained Preference Modeling

• Link：https://arxiv.org/abs/2603.01163
• Code：

owards Source-Aware Object Swapping with Initial Noise Perturbation

• Link：https://arxiv.org/abs/2602.23697
• Code：

Cycle-Consistent Tuning for Layered Image Decomposition

• Link：https://arxiv.org/abs/2602.20989
• Code：https://vcc.tech/research/2026/ImgDecom

ChordEdit: One-Step Low-Energy Transport for Image Editing

• Link：https://arxiv.org/abs/2602.19083
• Code：

Towards Source-Aware Object Swapping with Initial Noise Perturbation

• Link：https://arxiv.org/abs/2602.23697
• Code：

图像补全/图像修复(Image Inpainting)

1. HiFi-Inpaint: Towards High-Fidelity Reference-Based Inpainting for Generating Detail-Preserving Human-Product Images

• Link：https://arxiv.org/abs/2603.02210
• Code：https://correr-zhou.github.io/HiFi-Inpaint/

HiFi-Inpaint: Towards High-Fidelity Reference-Based Inpainting for Generating Detail-Preserving Human-Product Images

• Link：https://arxiv.org/abs/2603.02210
• Code：https://correr-zhou.github.io/HiFi-Inpaint/

生成对抗网络(GAN)

视频编辑(Video Editing)

Object-WIPER: Training-Free Object and Associated Effect Removal in Videos

• Link：https://arxiv.org/abs/2601.06391
• Code：

HorizonForge: Driving Scene Editing with Any Trajectories and Any Vehicles

• Link：https://arxiv.org/abs/2602.21333
• Code：https://horizonforge.github.io/

Low-level Vision

ShiftLUT: Spatial Shift Enhanced Look-Up Tables for Efficient Image Restoration

• Link：https://arxiv.org/abs/2603.00906
• Code：https://github.com/Sailor-t/ShiftLUT

F²HDR: Two-Stage HDR Video Reconstruction via Flow Adapter and Physical Motion Modeling

• Link：https://arxiv.org/abs/2603.14920
• Code：

Missing No More: Dictionary-Guided Cross-Modal Image Fusion under Missing Infrared

• Link：https://arxiv.org/abs/2603.08018
• Code：https://github.com/harukiv/DCMIF

BluRef: Unsupervised Image Deblurring with Dense-Matching References

• Link：https://arxiv.org/abs/2603.14176
• Code：https://qualcomm-ai-research.github.io/BluRef/

Towards Universal Computational Aberration Correction in Photographic Cameras: A Comprehensive Benchmark Analysis

• Link：https://arxiv.org/abs/2603.12083
• Code：https://github.com/XiaolongQian/UniCAC

Cross-Scale Pansharpening via ScaleFormer and the PanScale Benchmark

• Link：https://arxiv.org/abs/2603.00543
• Code：

ShiftLUT: Spatial Shift Enhanced Look-Up Tables for Efficient Image Restoration

• Link：https://arxiv.org/abs/2603.00906
• Code：

Reparameterized Tensor Ring Functional Decomposition for Multi-Dimensional Data Recovery

• Link：https://arxiv.org/abs/2603.01034
• Code：https://github.com/YangyangXu2002/RepTRFD

Lumosaic: Hyperspectral Video via Active Illumination and Coded-Exposure Pixels

• Link：https://arxiv.org/abs/2602.22140
• Code：

MatchED: Crisp Edge Detection Using End-to-End, Matching-based Supervision

• Link：https://arxiv.org/abs/2602.20689
• Code：https://cvpr26-matched.github.io

Continuous Exposure-Time Modeling for Realistic Atmospheric Turbulence Synthesis

• Link：https://arxiv.org/abs/2603.01398
• Code：https://github.com/Jun-Wei-Zeng/ET-Turb

超分辨率(Super-Resolution)

RAW-Domain Degradation Models for Realistic Smartphone Super-Resolution

• Link：https://arxiv.org/abs/2603.12493
• Code：

UCAN: Unified Convolutional Attention Network for Expansive Receptive Fields in Lightweight Super-Resolution

• Link：https://arxiv.org/abs/2603.11680
• Code：

FiDeSR: High-Fidelity and Detail-Preserving One-Step Diffusion Super-Resolution

• Link：https://arxiv.org/abs/2603.02692
• Code：https://github.com/Ar0Kim/FiDeSR

Toward Real-world Infrared Image Super-Resolution: A Unified Autoregressive Framework and Benchmark Dataset

• Link：https://arxiv.org/abs/2603.04745
• Code：https://github.com/JZD151/Real-IISR

AlignVAR: Towards Globally Consistent Visual Autoregression for Image Super-Resolution

• Link：https://arxiv.org/abs/2603.00589
• Code：

Spectral Super-Resolution via Adversarial Unfolding and Data-Driven Spectrum Regularization: From Multispectral Satellite Data to NASA Hyperspectral Image

• Link：https://arxiv.org/abs/2603.00920
• Code：https://sites.google.com/view/chiahsianglin/software

AlignVAR: Towards Globally Consistent Visual Autoregression for Image Super-Resolution

• Link：https://arxiv.org/abs/2603.00589
• Code：

去噪(Denoising)

Statistical Characteristic-Guided Denoising for Rapid High-Resolution Transmission Electron Microscopy Imaging

• Link：https://arxiv.org/abs/2603.18834
• Code：https://github.com/HeasonLee/SCGN

图像生成(Image Generation)

coDrawAgents: A Multi-Agent Dialogue Framework for Compositional Image Generation

• Link：https://arxiv.org/abs/2603.12829
• Code：

OmniLottie: Generating Vector Animations via Parameterized Lottie Tokens

• Link：https://arxiv.org/abs/2603.02138
• Code：https://openvglab.github.io/OmniLottie/

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

• Link：https://arxiv.org/abs/2603.00918
• Code：https://wookiekim.github.io/SOLACE/

VeCoR -- Velocity Contrastive Regularization for Flow Matching

• Link：https://arxiv.org/abs/2511.18942
• Code：https://p458732.github.io/VeCoR_Project_Page/

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

• Link：https://arxiv.org/abs/2603.00918
• Code：https://wookiekim.github.io/ARC/

Enhancing Spatial Understanding in Image Generation via Reward Modeling

• Link：https://arxiv.org/abs/2602.24233
• Code：https://github.com/DAGroup-PKU/SpatialT2I

AutoDebias: Automated Framework for Debiasing Text-to-Image Models

• Link：https://arxiv.org/abs/2508.00445
• Code：

视频生成(Video Generation)

Anchoring and Rescaling Attention for Semantically Coherent Inbetweening

• Link：https://arxiv.org/abs/2603.17651
• Code：https://github.com/teunchoi/TGI

Training-free Motion Factorization for Compositional Video Generation

• Link：https://arxiv.org/abs/2603.09104
• Code：

Chain of Event-Centric Causal Thought for Physically Plausible Video Generation

• Link：https://arxiv.org/abs/2603.09094
• Code：

FastLightGen: Fast and Light Video Generation with Fewer Steps and Parameters

• Link：https://arxiv.org/abs/2603.01685
• Code：

EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation

• Link：https://arxiv.org/abs/2603.12267
• Code：https://silentview.github.io/EVATok/

Geometry-as-context: Modulating Explicit 3D in Scene-consistent Video Generation to Geometry Context

• Link：https://arxiv.org/abs/2602.21929
• Code：

NOVA: Sparse Control, Dense Synthesis for Pair-Free Video Editing

• Link：https://arxiv.org/abs/2603.02802
• Code：

CubeComposer: Spatio-Temporal Autoregressive 4K 360° Video Generation from Perspective Video

• Link：https://arxiv.org/abs/2603.04291
• Code：https://lg-li.github.io/project/cubecomposer

FaceCam: Portrait Video Camera Control via Scale-Aware Conditioning

• Link：https://arxiv.org/abs/2603.05506
• Code：https://weijielyu.github.io/FaceCam

UniTalking: A Unified Audio-Video Framework for Talking Portrait Generation

• Link：https://arxiv.org/abs/2603.01418
• Code：

ExpPortrait: Expressive Portrait Generation via Personalized Representation

• Link：https://arxiv.org/abs/2602.19900
• Code：

LinVideo: A Post-Training Framework towards O(n) Attention in Efficient Video Generation

• Link：https://arxiv.org/abs/2510.08318
• Code：

FastLightGen: Fast and Light Video Generation with Fewer Steps and Parameters

• Link：https://arxiv.org/abs/2603.01685
• Code：无

Echoes Over Time: Unlocking Length Generalization in Video-to-Audio Generation Models

• Link：https://arxiv.org/abs/2602.20981
• Code：

The devil is in the details: Enhancing Video Virtual Try-On via Keyframe-Driven Details Injection

• Link：https://arxiv.org/abs/2512.20340
• Code：

3D生成

Edit-As-Act: Goal-Regressive Planning for Open-Vocabulary 3D Indoor Scene Editing

• Link：https://arxiv.org/abs/2603.17583
• Code：

NI-Tex: Non-isometric Image-based Garment Texture Generation

• Link：https://arxiv.org/abs/2511.18765
• Code：

ForgeDreamer: Industrial Text-to-3D Generation with Multi-Expert LoRA and Cross-View Hypergraph

• Link：https://arxiv.org/abs/2603.09266
• Code：

Zero-Shot Reconstruction of Animatable 3D Avatars with Cloth Dynamics from a Single Image

• Link：https://arxiv.org/abs/2603.14772
• Code：

BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation

• Link：https://arxiv.org/abs/2602.18873
• Code：https://wangmiaowei.github.io/BiMotion.github.io/

Mesh-Pro: Asynchronous Advantage-guided Ranking Preference Optimization for Artist-style Quadrilateral Mesh Generation

• Link：https://arxiv.org/abs/2603.00526
• Code：

MorphAny3D: Unleashing the Power of Structured Latent in 3D Morphing

• Link：https://arxiv.org/abs/2601.00204
• Code：https://xiaokunsun.github.io/MorphAny3D.github.io

Easy3E: Feed-Forward 3D Asset Editing via Rectified Voxel Flow

• Link：https://arxiv.org/abs/2602.21499
• Code：

BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation

• Link：https://arxiv.org/abs/2602.18873
• Code：

视频理解(Video Understanding)

F2HDR: Two-Stage HDR Video Reconstruction via Flow Adapter and Physical Motion Modeling

• Link：https://arxiv.org/abs/2603.14920
• Code：

Training-free Detection of Generated Videos via Spatial-Temporal Likelihoods

• Link：https://arxiv.org/abs/2603.15026
• Code：https://omerbenhayun.github.io/stall-video

PFGNet: A Fully Convolutional Frequency-Guided Peripheral Gating Network for Efficient Spatiotemporal Predictive Learning

• Link：https://arxiv.org/abs/2602.20537
• Code：https://github.com/fhjdqaq/PFGNet

Wavelet-based Frame Selection by Detecting Semantic Boundary for Long Video Understanding

• Link：https://arxiv.org/abs/2603.00512
• Code：https://github.com/MAC-AutoML/WFS-SB

Question-guided Visual Compression with Memory Feedback for Long-Term Video Understanding

• Link：https://arxiv.org/abs/2603.15167
• Code：

StreamingTOM: Streaming Token Compression for Efficient Video Understanding

• Link：https://arxiv.org/abs/2510.18269
• Code：https://yige24.github.io/StreamingTOM

Follow the Saliency: Supervised Saliency for Retrieval-augmented Dense Video Captioning

• Link：https://arxiv.org/abs/2603.11460
• Code：https://github.com/ermitaju1/STaRC

Stay in your Lane: Role Specific Queries with Overlap Suppression Loss for Dense Video Captioning

• Link：https://arxiv.org/abs/2603.11439
• Code：

VirtueBench: Evaluating Trustworthiness under Uncertainty in Long Video Understanding

• Link：https://arxiv.org/abs/2603.07071
• Code：

StreamReady: Learning What to Answer and When in Long Streaming Videos

• Link：https://arxiv.org/abs/2603.08620
• Code：

SAIL: Similarity-Aware Guidance and Inter-Caption Augmentation-based Learning for Weakly-Supervised Dense Video Captioning

• Link：https://arxiv.org/abs/2603.05437
• Code：

LongVideo-R1: Smart Navigation for Low-cost Long Video Understanding

• Link：https://arxiv.org/abs/2602.20913
• Code：https://github.com/qiujihao19/LongVideo-R1

VideoChat-M1: Collaborative Policy Planning for Video Understanding via Multi-Agent Reinforcement Learning

• Link：https://arxiv.org/abs/2511.19524
• Code：

SAIL: Similarity-Aware Guidance and Inter-Caption Augmentation-based Learning for Weakly-Supervised Dense Video Captioning

• Link：https://arxiv.org/abs/2603.05437
• Code：

Think, Then Verify: A Hypothesis-Verification Multi-Agent Framework for Long Video Understanding

• Link：https://arxiv.org/abs/2603.04977
• Code：https://github.com/Haorane/VideoHV-Agent

ViterbiPlanNet: Injecting Procedural Knowledge via Differentiable Viterbi for Planning in Instructional Videos

• Link：https://arxiv.org/abs/2603.04265
• Code：

Wavelet-based Frame Selection by Detecting Semantic Boundary for Long Video Understanding

• Link：https://arxiv.org/abs/2603.00512
• Code：

Exploring Spatiotemporal Feature Propagation for Video-Level Compressive Spectral Reconstruction: Dataset, Model and Benchmark

• Link：https://arxiv.org/abs/2603.00611
• Code：https://github.com/nju-cite/DynaSpec

Frame2Freq: Spectral Adapters for Fine-Grained Video Understanding

• Link：https://arxiv.org/abs/2602.18977
• Code：https://github.com/th-nesh/Frame2Freq

FluxMem: Adaptive Hierarchical Memory for Streaming Video Understanding

• Link：https://arxiv.org/abs/2603.02096
• Code：https://yiwengxie.com/FluxMem/

Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models

• Link：https://arxiv.org/abs/2603.01400
• Code：https://tyroneli.github.io/AOT

3D人体姿态估计(3D Human Pose Estimation)

GazeOnce360: Fisheye-Based 360° Multi-Person Gaze Estimation with Global-Local Feature Fusion

• Link：https://arxiv.org/abs/2603.17161
• Code：https://caizhuojiang.github.io/GazeOnce360/

OnlineHMR: Video-based Online World-Grounded Human Mesh Recovery

• Link：https://arxiv.org/abs/2603.17355
• Code：https://tsukasane.github.io/Video-OnlineHMR/

Shoe Style-Invariant and Ground-Aware Learning for Dense Foot Contact Estimation

• Link：https://arxiv.org/abs/2511.22184
• Code：https://github.com/dqj5182/FECO_RELEASE

Towards Balanced Multi-Modal Learning in 3D Human Pose Estimation

• Link：https://arxiv.org/abs/2501.05264
• Code：https://github.com/MICLAB-BUPT/AWC

Enhancing Hands in 3D Whole-Body Pose Estimation with Conditional Hands Modulator

• Link：https://arxiv.org/abs/2603.14726
• Code：

CIGPose: Causal Intervention Graph Neural Network for Whole-Body Pose Estimation

• Link：https://arxiv.org/abs/2603.09418
• Code：https://github.com/53mins/CIGPose

EgoPoseFormer v2: Accurate Egocentric Human Motion Estimation for AR/VR

• Link：https://arxiv.org/abs/2603.04090
• Code：

Sketch2Colab: Sketch-Conditioned Multi-Human Animation via Controllable Flow Distillation

• Link：https://arxiv.org/abs/2603.02190
• Code：

SIMSPINE: A Biomechanics-Aware Simulation Framework for 3D Spine Motion Annotation and Benchmarking

• Link：https://arxiv.org/abs/2602.20792
• Code：

VLM-Guided Group Preference Alignment for Diffusion-based Human Mesh Recovery

• Link：https://arxiv.org/abs/2602.19180
• Code：

持续学习(Continual Learning)

Lifelong Imitation Learning with Multimodal Latent Replay and Incremental Adjustment

• Link：https://arxiv.org/abs/2603.10929
• Code：https://github.com/yfqi/lifelong_mlr_ifa

Elastic Weight Consolidation Done Right for Continual Learning

• Link：https://arxiv.org/abs/2603.18596
• Code：

行为识别(Action Recognition)

BriMA: Bridged Modality Adaptation for Multi-Modal Continual Action Quality Assessment

• Link：https://arxiv.org/abs/2602.19170
• Code：

Test-time Ego-Exo-centric Adaptation for Action Anticipation via Multi-Label Prototype Growing and Dual-Clue Consistency

• Link：https://arxiv.org/abs/2603.09798
• Code：https://github.com/ZhaofengSHI/DCPGN

知识蒸馏(Knowledge Distillation)

Momentum Memory for Knowledge Distillation in Computational Pathology

• Link：https://arxiv.org/abs/2602.21395
• Code：

WaDi: Weight Direction-aware Distillation for One-step Image Synthesis

• Link：https://arxiv.org/abs/2603.08258
• Code：https://github.com/gudaochangsheng/WaDi

Fixed Anchors Are Not Enough: Dynamic Retrieval and Persistent Homology for Dataset Distillation

• Link：https://arxiv.org/abs/2602.24144
• Code：

Brewing Stronger Features: Dual-Teacher Distillation for Multispectral Earth Observation

• Link：https://arxiv.org/abs/2602.19863
• Code：https://wolfilip.github.io/DEO/

Distilling Balanced Knowledge from a Biased Teacher

• Link：https://arxiv.org/abs/2506.18496
• Code：

Momentum Memory for Knowledge Distillation in Computational Pathology

• Link：https://arxiv.org/abs/2602.21395
• Code：https://github.com/CAIR-LAB-WFUSM/MoMKD

图像压缩(Image Compression)

SGI: Structured 2D Gaussians for Efficient and Compact Large Image Representation

• Link：https://arxiv.org/abs/2603.07789
• Code：https://github.com/zx-pan/SGI

Parallax to Align Them All: An OmniParallax Attention Mechanism for Distributed Multi-View Image Compression

• Link：https://arxiv.org/abs/2603.03615
• Code：

Zero-Shot Learning(零样本学习)

Learning through Creation: A Hash-Free Framework for On-the-Fly Category Discovery

• Link：https://arxiv.org/abs/2603.13858
• Code：https://github.com/brandinzhang/LTC

TALON: Test-time Adaptive Learning for On-the-Fly Category Discovery

• Link：https://arxiv.org/abs/2603.08075
• Code：https://github.com/ynanwu/TALON

Learning through Creation: A Hash-Free Framework for On-the-Fly Category Discovery

• Link：https://arxiv.org/abs/2603.13858
• Code：https://github.com/brandinzhang/LTC

Boosting Quantitive and Spatial Awareness for Zero-Shot Object Counting

• Link：https://arxiv.org/abs/2603.16129
• Code：

立体匹配(Stereo Matching)

PromptStereo: Zero-Shot Stereo Matching via Structure and Motion Prompts

• Link：https://arxiv.org/abs/2603.01650
• Code：

PromptStereo: Zero-Shot Stereo Matching via Structure and Motion Prompts

• Link：https://arxiv.org/abs/2603.01650
• Code：

Pip-Stereo: Progressive Iterations Pruner for Iterative Optimization based Stereo Matching

• Link：https://arxiv.org/abs/2602.19112
• Code：https://github.com/XPENG-Aridge-AI

场景图生成(Scene Graph Generation)

DSFlash: Comprehensive Panoptic Scene Graph Generation in Realtime

• Link：https://arxiv.org/abs/2603.10538
• Code：

计数(Counting)

UNICBench: UNIfied Counting Benchmark for MLLM

• Link：https://arxiv.org/abs/2603.00595
• Code：

隐式神经表示(Implicit Neural Representations)

图像质量评价(Image Quality Assessment)

How to Take a Memorable Picture? Empowering Users with Actionable Feedback

• Link：https://arxiv.org/abs/2602.21877
• Code：https://laitifranz.github.io/MemCoach/

Fine-grained Image Aesthetic Assessment: Learning Discriminative Scores from Relative Ranks

• Link：https://arxiv.org/abs/2603.03907
• Code：

视频质量评价(Video Quality Assessment)

数据集(Datasets)

E-comIQ-ZH: A Human-Aligned Dataset and Benchmark for Fine-Grained Evaluation of E-commerce Posters with Chain-of-Thought

• Link：https://arxiv.org/abs/2602.21698
• Code：https://github.com/4mm7/E-comIQ-ZH

Continuous Exposure-Time Modeling for Realistic Atmospheric Turbulence Synthesis

• Link：https://arxiv.org/abs/2603.01398
• Code：https://github.com/Jun-Wei-Zeng/ET-Turb

LenghuSky-8: An 8-Year All-Sky Cloud Dataset with Star-Aware Masks and Alt-Az Calibration for Segmentation and Nowcasting

• Link：https://arxiv.org/abs/2603.16429
• Code：

AVA-Bench: Atomic Visual Ability Benchmark for Vision Foundation Models

• Link：https://arxiv.org/abs/2506.09082
• Code：

反学习(Machine Unlearning)

SineProject: Machine Unlearning for Stable Vision Language Alignment

• Link：https://arxiv.org/abs/2511.18444
• Code：

RAZOR: Ratio-Aware Layer Editing for Targeted Unlearning in Vision Transformers and Diffusion Models

• Link：https://arxiv.org/abs/2603.14819
• Code：

Stake the Points: Structure-Faithful Instance Unlearning

• Link：https://arxiv.org/abs/2603.12915
• Code：

新任务(New Tasks)

模型加速(Improving Reasoning)

Flash-Unified: A Training-Free and Task-Aware Acceleration Framework for Native Unified Models

• Link：https://arxiv.org/abs/2603.15271
• Code：https://github.com/Rirayh/FlashU

Variation-aware Vision Token Dropping for Faster Large Vision-Language Models

• Link：https://arxiv.org/abs/2509.01552
• Code：https://github.com/xuyang-liu16/V2Drop

Model Merging in the Essential Subspace

• Link：https://arxiv.org/abs/2602.20208
• Code：

时间序列(Time Series)

STCast: Adaptive Boundary Alignment for Global and Regional Weather Forecasting

• Link：https://arxiv.org/abs/2509.25210
• Code：

脉冲网络

Stable Spike: Dual Consistency Optimization via Bitwise AND Operations for Spiking Neural Networks

• Link：https://arxiv.org/abs/2603.11676
• Code：

Rethinking SNN Online Training and Deployment: Gradient-Coherent Learning via Hybrid-Driven LIF Model

• Link：https://arxiv.org/abs/2410.07547
• Code：https://github.com/hzc1208/HD_LIF

图像检索

TIACam: Text-Anchored Invariant Feature Learning with Auto-Augmentation for Camera-Robust Zero-Watermarking

• Link：https://arxiv.org/abs/2602.18863
• Code：

PinPoint: Evaluation of Composed Image Retrieval with Explicit Negatives, Multi-Image Queries, and Paraphrase Testing

• Link：https://arxiv.org/abs/2603.04598
• Code：

NaiLIA: Multimodal Nail Design Retrieval Based on Dense Intent Descriptions and Palette Queries

• Link：https://arxiv.org/abs/2603.05446
• Code：

其他(Others)

NESTOR: A Nested MOE-based Neural Operator for Large-Scale PDE Pre-Training

• Link：https://arxiv.org/abs/2602.22059
• Code：

Defending Unauthorized Model Merging via Dual-Stage Weight Protection

• Link：https://arxiv.org/abs/2511.11851
• Code：

BD-Merging: Bias-Aware Dynamic Model Merging with Evidence-Guided Contrastive Learning

• Link：https://arxiv.org/abs/2603.03920
• Code：

ACE-Merging: Data-Free Model Merging with Adaptive Covariance Estimation

• Link：https://arxiv.org/abs/2603.02945
• Code：

GEM-TFL: Bridging Weak and Full Supervision for Forgery Localization through EM-Guided Decomposition and Temporal Refinement

• Link：https://arxiv.org/abs/2603.05095
• Code：

DC-Merge: Improving Model Merging with Directional Consistency

• Link：https://arxiv.org/abs/2603.06242
• Code：https://github.com/Tobeginwith/DC-Merge

Defending Unauthorized Model Merging via Dual-Stage Weight Protection

• Link：https://arxiv.org/abs/2511.11851
• Code：

BD-Merging: Bias-Aware Dynamic Model Merging with Evidence-Guided Contrastive Learning

• Link：https://arxiv.org/abs/2603.03920
• Code：

Bridging Domains through Subspace-Aware Model Merging

• Link：https://arxiv.org/abs/2603.05768
• Code：