跳转至

VLM 前沿增量 · 2026-10

生成：py -3.9 tools/watch.py --volume vlm --write｜窗口 2026-06-11 ~ 2026-10-09（120 天）｜词表 24 条｜请求失败 0 条｜新增 449 条

词表按 docs/04-VLM专题/05–07 三篇 分节。

读法（纪律 F 的镜像规则）：TOTAL 是题摘层面的命中数，不是相关论文数。一条宽查询命中几百条不含任何信息，所以下表只列仓库尚未收录的条目。NULL 表示请求失败（429/超时），不是 0 命中；0 表示确实没有，但第一反应应当是怀疑查询措辞，换个提法复核后再下结论。

arXiv 的 all: / abs: 只覆盖题名/摘要/作者/注释，不覆盖全文。

去重范围是整仓：docs/、references/、mindmaps/、paper/ 下的 .md，加 README.md / CONTRIBUTING.md / code/README.md。paper/ 下的 _work*/ 取数缓存（json/xml/html/pdf）不算收录，故不计入——那里放的是原始查询结果，不是读过的文献。所以某条 ID 不在本表，只能推出整仓没写过，推不出"这个专题没有"。 （2026-10-08 修：paper/ 原先漏扫，加上 ID 正则不吃 v1 版本号后缀，两条合起来让 27 条早已在 paper/ 笔记里读过的论文被标成"新增"。三卷的台账互为已收录，故 <卷>-watch-*.md 一律排除。）

05 · 通用架构与视觉编码器

abs:"vision encoder" AND abs:"multimodal large language" — 命中 17，其中新增 17

日期 arXiv 标题
2026-10-07 2610.09440v1 Mixture of Layers: Dynamic Layer Routing for Visual Reasoning
2026-10-04 2610.05413v1 A Strong Baseline for Evaluating Vision Encoders in Multimodal Large Language Models
2026-09-30 2609.40361v1 Ranking-Aware Prompt Optimization for Multimodal Clinical Diagnosis
2026-09-28 2609.34977v1 SPIDER: Multi-Layer Semantic Token Pruning and Adaptive Sub-Layer Skipping in Multimodal Large Language Models
2026-09-07 2609.07823v1 SAFIRE: Safety-Critical Benchmark for Fine-grained Fire and Smoke Understanding in Multimodal LLMs
2026-08-30 2608.29802v1 Foundation and Multimodal Large Language Models for Face Presentation and Morph Attack Detection
2026-08-28 2608.28312v1 AIM: Anchor Identity Features, Then Match for Multimodal Large Language Model Unlearning
2026-07-27 2607.24743v2 ClinFusion: A Vision-Centric Multimodal LLM System for Holistic Medical Understanding
2026-07-20 2607.17806v1 PGN: Design and Implementation of a Vision-Language Navigation System Based on Pangu Multimodal Foundation Mod
2026-07-13 2607.12112v2 Continual Learning with Elastic Regularization and Synthetic Replay for Federated MLLM Fine-Tuning
2026-07-13 2607.11562v1 MonkeyOCRv2: A Visual-Text Foundation Model for Document AI
2026-07-08 2607.07673v1 MedPMC: A Systematic Framework for Scaling High-Fidelity Medical Multimodal Data for Foundation Models
2026-07-07 2607.05880v1 Harrison.Rad 1.5 Technical Report: A radiology foundation model that can draft reports from images, priors and
2026-07-01 2607.00416v1 DroneIQA-VLE: Multi-Task Drone Image Quality Assessment via Vision-Language Ensemble
2026-06-23 2606.25084v1 Are We There Yet? Exploring the Capabilities of MLLMs in Assistive AI Applications
2026-06-22 2606.23885v1 Mind the Heads: Topological Representation Alignment for Multimodal LLMs
2026-06-13 2606.15134v1 Beyond Scalar Distances: Semantic Attribute Gradients from Frozen MLLMs for Visual Embeddings

abs:"connector" AND abs:"multimodal large language" — 命中 1，其中新增 1

日期 arXiv 标题
2026-09-18 2609.21675v1 DRT: Dense Reasoning Trace for Efficient and Grounded Multimodal Reasoning

abs:"SigLIP" — 命中 43，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-01 2610.02148v1 Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation
2026-09-30 2609.39625v1 D-Scope: Decomposing and Steering Diffusion Transformers with Sparse Autoencoders
2026-09-29 2609.37685v2 PAIQ: Patch-Aligned Semantic Injection via Residual Rotation
2026-09-28 2609.36101v1 One Geometry, Different Outcomes: Readout-Dependent Effects of the Modality Gap in Vision-Language Models
2026-09-28 2609.35470v1 Representation Risk in Pretrained Image Encoders
2026-09-27 2609.34021v1 Position Aware Layer Queries for Test Time Training in Vision Language Models
2026-09-27 2609.33359v1 When Does Geometric View Synthesis Help Wine Label Retrieval? A Public One-Shot Benchmark Across Self-Supervis
2026-09-23 2609.28857v1 MEVL-STP: Multi-Encoder and Vision Language Model for Arbitrarily Shaped Scene Text Spotting
2026-09-22 2609.27142v1 MINER: Multi-crop INference-time Enhancement for Rare-Object Retrieval with Frozen Dual Encoders
2026-09-09 2609.10224v1 UOT-Gap: A Variational Principle for the Modality Gap in Vision-Language Models via Unbalanced Optimal Transpo
2026-09-07 2609.07937v1 TDDN: Text-aligned Diffused DINO Network for Puzzle Understanding
2026-09-02 2609.03085v2 Solving the Needle-in-a-Haystack Problem in Mammography Vision-Language Model with Differentiable Subset Sampl
2026-08-30 2608.29974v1 SpanCalib-VLM: Calibrated Hallucination Span Detection in Vision-Language Models
2026-08-29 2608.29395v1 GATE: Reliability-Gated Gaussian Evidence Fusion for Training-Free Test-Time Adaptation of Vision-Language Mod
2026-08-28 2608.28316v2 Conditional Visual Evidence Utility: State-Dependent Rank Reversals in Frozen Vision-Language Encoders
2026-08-25 2608.24133v1 PlaceSeek: Human-Centered Geospatial Retrieval of Urban Outdoor Places via Semantic Grounding and Affective Al
2026-08-24 2608.23484v1 Multi-Modal Semantic Expansion with Constrained LLM Reranking for Conversational Music Recommendation
2026-08-23 2608.23634v1 The Blending Ratio Is Not Where the Performance Is: Diagnosing Prototype Blending for Few-Shot Adaptation of V
2026-08-21 2608.20988v1 Jacobian-guided Noise Injection for Quantization Robustness in Large Language Models
2026-08-19 2608.19376v1 Does Marginal Coverage Guarantee Class-Conditional Safety for Zero-Shot VLMs Under Shift?
2026-08-19 2608.21443v1 Text-Guided Visual Dependency Graph Learning with Cross-Modal Attention Priors
2026-08-19 2608.18591v1 Can a Lightweight Multimodal Model Estimate LLM Reasoning Performance? A Study for Compute-Optimal Document In
2026-08-17 2608.16690v2 AnchorScore: A CLIP-Based Diagnostic of MLLM Annotation Difficulty
2026-08-17 2608.16975v1 Margin-Regularized Structured Semantic Alignment for Brain-Language Correspondence
2026-08-17 2608.16263v1 Seeing Before Answering: Training-Free Visual Layer Profiling for Vision-Language Models
2026-08-17 2608.16234v1 GaussianDWM++: Language-Grounded 3D Gaussian Driving World Model for Unified Scene Understanding, Editing, and
2026-08-17 2608.16962v1 Clinical Pathways Matter for Multimodal Deep Learning in Early Alzheimers Disease Detection
2026-08-11 2609.26168v1 TRACE: Transparent Retrieval for Abstract Concept Evaluation
2026-08-09 2608.08477v5 VectraYX-Vision-1B: A Sub-2B Spanish/LATAM Cybersecurity Vision-Language Model, and What Limits Its Visual Gro
2026-07-28 2607.25842v1 Adversarial Deepfake Generation and an Investigation of Purification-Based Adversarial Detection

abs:"visual token" AND abs:"multimodal" — 命中 144，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-06 2610.07984v1 Decide Before You Look: Learning Which Retrieved Memories Deserve Pixels
2026-10-06 2610.07572v1 Two Vectors Replace In-Context Demos: Structured Task Adaptation via Embeddings
2026-10-04 2610.05097v1 ReMAP: Restoring the Perceptual Cycle with Reasoning-Time Latent Visual Memory
2026-10-03 2610.04225v1 FlashGaze: Training-Free Multi-Scale Patch Pruning For Efficient Video Understanding
2026-10-02 2610.03400v1 Beyond Entropy: Self-Diagnostic Multi-Role Token Optimization for Video Reasoning
2026-10-02 2610.03822v1 Beyond Token Accuracy: Prioritizing What Matters for Visual Reconstruction
2026-09-30 2610.00623v1 HAWK: Rethinking Multimodal Drafting for Speculative Decoding
2026-09-30 2609.39924v1 CoVisco: Codec-Native Vision Encoder with Native Token Compression for Unified Image-Video Understanding
2026-09-30 2609.38823v1 DecoMoE: Decoupling Visual Propagation and Expert Computation for Efficient Multimodal MoE Inference
2026-09-29 2609.38485v1 Beyond Layers: Position-Resolved Gradient Conflict and Position-Aware Modulation for Unified Multimodal Models
2026-09-29 2609.37225v1 ResComEmb: Effective and Efficient Multimodal Embedding via Residual Homogeneity Compression
2026-09-29 2609.37096v2 Why MLLMs Struggle to Count: Overcoming Individuation and Aggregation Bottlenecks with ConvStack
2026-09-29 2609.37052v1 OmniRoute: Mapping Temporal Semantic Evidence to Audio-Visual Token Budgets for Efficient Omnimodal Large Lang
2026-09-29 2609.36916v2 Representation Dynamics Reveal Semantic Saliency and Similarity for Visual Token Pruning in MLLMs
2026-09-29 2609.36782v1 Decoding Affective Nuances: Enhancing MLLMs via Hierarchical Emotion Reasoning and Contrastive Discriminative
2026-09-28 2609.36145v1 From Sharp Eyes to Expert Mind: Internalizing Expert Knowledge in MLLMs for Tampered Text Detection
2026-09-28 2609.35457v1 How Far Are We from Removing the Visual Encoder? Scaling Laws for Encoder-Free Multimodal Pretraining
2026-09-28 2609.34977v1 SPIDER: Multi-Layer Semantic Token Pruning and Adaptive Sub-Layer Skipping in Multimodal Large Language Models
2026-09-28 2609.34972v1 Just MLPs: Efficient Visual State Reconstruction for Multimodal Language Models
2026-09-28 2609.34942v1 Resolution as a First-Class Decision: Task-Conditioned Routing for Efficient Multimodal Large Language Models
2026-09-28 2609.34867v1 P4Q: Co-designing Token Pruning and Quantization for Vision-Language Model Acceleration
2026-09-28 2609.34408v1 Distilling Visual Reasoning into Text Space
2026-09-28 2609.34330v2 MiCo: Mutual Information Coverage Optimization through Semantic Erasure Modeling for Efficient MLLM Inference
2026-09-28 2609.34196v1 ConvCue: Complementary Visual Inductive Biases for Vision-Language Models
2026-09-27 2609.33518v1 SceneScaffold: Active Scene-State Construction for Unified 3D Scene Understanding
2026-09-26 2609.32353v2 Fewer Tokens, More Self-Teaching: On-Policy Self-Distillation for Extreme Visual Token Reduction
2026-09-26 2609.32333v1 Progressive-View On-Policy Distillation for Regional-to-Global Transfer in Multimodal LLMs
2026-09-25 2609.30783v1 Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Mo
2026-09-24 2609.30210v1 The Alignment Illusion in Multimodal Large Language Models
2026-09-24 2609.29940v1 Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration

abs:"native resolution" AND abs:"vision-language" — 命中 5，其中新增 5

日期 arXiv 标题
2026-10-06 2610.07729v1 Foveated Compression: Selective High-Resolution Preservation for Token-Efficient VLMs
2026-10-03 2610.04318v1 Rethinking Long-Video Efficiency: A Joint Allocation Perspective on Frames, Pixels, and Front-End Latency
2026-09-18 2609.21712v2 ZYT-World: A Real-Time Controllable World Model for Closed-Loop Autonomous-Driving Simulation
2026-09-09 2609.13287v1 LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents
2026-06-18 2606.20523v2 SARLO-80: Worldwide Slant SAR Language Optic Dataset 80cm

abs:"contrastive" AND abs:"vision-language" AND abs:"pretraining" — 命中 42，其中新增 29（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-09-30 2609.39924v1 CoVisco: Codec-Native Vision Encoder with Native Token Compression for Unified Image-Video Understanding
2026-09-30 2609.38989v1 Cue the Flow: Steering Flow-Matching Policies for Open-World Delivery Manipulation
2026-09-28 2609.34944v1 Adjoint Guidance Flow: Amortized Critic Guidance for VLA Policies
2026-09-28 2609.34826v1 WM-VLM: Probing Internal World Models for Interleaved Visual-Textual Reasoning
2026-09-28 2609.34479v1 SentZero: An Enhanced Sentence-Centric Vision-Language Pretraining for Multi-Task Zero-Shot Chest X-Ray Analys
2026-09-28 2609.34206v1 WorldGuide: Learning Success-Failure Boundaries in Latent World Models for Vision-Language-Action Policies
2026-09-26 2609.32582v1 Cool Embeddings: Predicting Galaxy Cluster Cooling Times in IllustrisTNG and TNG-Cluster with AstroCLIP
2026-09-26 2609.32108v1 SAMBAR: Selective Anchoring via Method of Multipliers for Balanced Knowledge Acquisition and Retention in Visi
2026-09-25 2609.31985v1 Does Vision-Language Pretraining Granularity Matter? A Controlled Evaluation of Vision-Language Objectives Acr
2026-09-24 2609.29156v1 Med-AR: Autoregressive Vision-Language Pretraining for Long-Tailed Chest X-Ray Classification and Uncertainty-
2026-09-17 2609.20648v2 SkipVLA: Skipping VLA Steps with Classical Planning for Fast Robot Manipulation
2026-09-15 2609.16664v1 Bridging the Perceptual Gap: Residual-Enhanced Downscaling and Manifold-Aware Perception Alignment Adaptation
2026-09-11 2609.12454v1 Bridging Vision Foundation Model Priors with CLIP for Spatial-aware Few-shot Anomaly Detection in Medical Imag
2026-09-10 2609.12035v2 Reading the Whole Heart: Latent-Attention Masked Autoencoders for Multimodal Cardiac Representation Learning
2026-09-04 2609.05583v1 An overview of 3D Vision-Language Models
2026-09-03 2609.03391v2 Exploring the Potential of Contrastive Language-Image Pre-training for Multi-Source Remote Sensing Data
2026-09-02 2609.03085v2 Solving the Needle-in-a-Haystack Problem in Mammography Vision-Language Model with Differentiable Subset Sampl
2026-09-01 2609.01757v1 AlphaRAD: Grounded Zero-Shot Classification in Chest Radiology via $α$-Corrected Binary Cross Entropy and Fact
2026-08-31 2609.00411v1 Unmasking Face Embeddings: Reading, Rendering and Naming with Foundation Models
2026-08-26 2608.25701v1 Skeleton-based Zero-Shot Spatio-Temporal Action Localization via Weakly-Supervised Pretraining
2026-08-16 2608.15456v1 AlignJEPA: Predictive Vision-Language Alignment for Remote Sensing Foundation Models
2026-08-11 2609.26168v1 TRACE: Transparent Retrieval for Abstract Concept Evaluation
2026-08-06 2608.05960v1 Big, Bright, or Invisible: A Frozen-Feature Benchmark of 3D CT Foundation Models
2026-08-05 2608.05389v1 Text-Guided Refinement of Multi-sequence Glioma Subregion Segmentation with a Vision-Language Foundation Model
2026-07-31 2608.00279v1 Learning to See Locally and Align Clinically with Pathology Semantics for Radiology Report Generation
2026-07-31 2608.00231v2 Learning How Much, Not Just What: Cross-Patient Burden Order for CT Vision-Language Pretraining
2026-07-29 2609.17572v1 Disentangling Algorithmic Bias from Archival Artifacts: A Controlled Audit of Vision-Language Model Valuation
2026-07-25 2607.23271v1 What CLIP Knows but Cannot Say: Recovering Negation from Frozen Intermediate Features
2026-07-24 2607.21970v1 TextSLIP: Text Self-Supervised CLIP for Medical Report Generation

06 · 指令微调与对齐

abs:"visual instruction tuning" — 命中 8，其中新增 8

日期 arXiv 标题
2026-09-05 2609.05885v1 One Rate Is Not Enough: Adaptive Anisotropic Learning Rates for LoRA Fine-Tuning
2026-08-31 2608.30209v3 DICS: Exploring Data Intrinsic Consistency for Visual Instruction Selection
2026-08-27 2608.26820v1 LLaVAFlow: Preserving Latent Alignment Flow for Parameter-Efficient Multimodal Fine-Tuning
2026-08-10 2608.09907v1 DistMoE: Private-data Rehearsal-free Routing in Mixture-of-Experts for Distributed Instruction Tuning
2026-08-03 2608.01635v1 Mitigating Visual Degradation in MLLMs via Spatial-Spectral Visual Anchor Learning
2026-07-29 2607.26596v1 Decoupled Visual Processing: Efficient Multimodal Adaptation via Modality-Specific Transformer Substitution
2026-07-01 2607.00465v1 StochasT: Learning with Stochastic Turn Depth for Visual Instruction Tuning
2026-06-17 2606.19100v5 AMALIA-VL: A Native European Portuguese Open-Source Vision and Language Model

abs:"LoRA" AND abs:"vision-language" — 命中 86，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-07 2610.10163v1 Beyond Anonymous Captions: Grounding Character Identity in Video Captioning and Question Answering
2026-10-07 2610.09940v1 Juno: Taming Predictive Latents for Vision-Language-Action Models
2026-10-05 2610.06184v1 Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models
2026-10-01 2610.02388v1 Octrees as an Explicit 3D Language
2026-10-01 2610.02148v1 Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation
2026-09-30 2609.39820v1 Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models
2026-09-29 2609.38391v1 Team MSU GenText-Forensics Challenge 2026 Technical Report
2026-09-29 2609.36492v1 Benchmarking Vision-Language Models on Synapse Detection and Proofreading in Connectomics
2026-09-28 2609.34968v1 RoboFL: Federated Expert Assembly for World Action Models
2026-09-28 2609.34561v1 Brain-Conditioned Action Policies for Neural Motor Decoding
2026-09-27 2609.33694v1 Seeing and Solving Are Not Enough for Vision-Language Models
2026-09-27 2609.33158v1 FOCUS: Benchmarking Retinal Model Generalization from Foundation Vision Encoders to Multimodal LLMs
2026-09-27 2609.33125v1 Train Together or Merge Later? Unifying VLA Experts via a Shared Action Interface
2026-09-27 2609.33082v1 SemReward-VL: Semantic Reward-Guided Video-Language Adaptation for Developmental Behavior Assessment
2026-09-25 2609.31450v1 From Reward Signal to Visual Utility: A Controlled Audit of Medical VLM Post-Training
2026-09-21 2609.24334v1 Multitask Jet Analysis with Vision-Language Models: A Physics-Informed Four-Panel Representation
2026-09-21 2609.24198v1 SKstars at SHROOM: Visions Agreement-Guided Ensembling of Zero-Shot and LoRA-Adapted Vision--Language Models
2026-09-20 2609.23655v1 Reassessing Global Gradient-Norm Imbalance in BLIP Fine-Tuning Across Physical Domains
2026-09-19 2609.23012v1 CrowdCue: Specialist-Cue Conditioning for Vision-Language Crowd Counting
2026-09-19 2609.22925v1 "Dear LLaVA, Please Drive": A Depth-Aware Vision-Language Agent for Closed-Loop Robotic Control
2026-09-16 2609.18860v2 Decodable but Misrouted: Sparse Features Uncover a Readout Gap in Vision-Language Models for Harmful Meme Dete
2026-09-16 2609.18084v2 Not All Layers Need Tuning: Diagnosing and Directing Adaptation in Vision-Language-Action Models
2026-09-15 2609.16486v1 VPRef: A Cross-Domain Benchmark for Referring Remote Sensing Image Segmentation
2026-09-14 2609.15603v1 A Unified Vision-Language Model for PSMA PET/CT Report Generation, Visual Question Answering, and Lesion Segme
2026-09-14 2609.15229v1 Pre-PEFT Probing: Weight Statistics and Perturbation Robustness for Layer Selection in VLM Vision Encoders
2026-09-13 2609.14523v1 Selective Tool Use for Agentic Change Visual Question Answering in Remote Sensing
2026-09-13 2609.14350v1 Two-Stage Mixture-of-LoRA for Multi-Task Medical Vision-Language Learning
2026-09-13 2609.14284v2 Vision-Language Models for Criterion-Level Grading of Handwritten Examinations in Outcome-Based Education
2026-09-10 2609.11310v1 Your Model Already Knows Don't Teach It, Learn to Ask It: Soft Prompting for Few-Shot Adaptation of Vision-Lan
2026-09-08 2609.08188v1 Bridging the Semantic-Utility Gap in Multimodal RAG via Generator-in-the-Loop Alignment

abs:"instruction tuning" AND abs:"multimodal" — 命中 49，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-09-29 2609.37889v1 ReCAP: Retrieval-Guided Capability Reuse for Multimodal Continual Instruction Tuning
2026-09-29 2609.37283v1 SAM Meets VLM: Parameter-Decoupled Full-Parameter Training for Unified Medical Reasoning and Segmentation
2026-09-28 2609.36121v1 Render Before Reading: Visual Rendering as a Prompt Injection Defense
2026-09-17 2609.20325v1 AgriScope: Pixel-Grounded Multimodal Understanding for Agricultural Images
2026-09-12 2609.13742v1 Hyper-LLaVA: Hyperbolic Uncertainty-aware Modality-Balanced Routing for Multimodal Continual Instruction Tunin
2026-09-01 2609.00654v1 SciTrue: Reliable Scientific Claim Validation with Frontier and Open Language Models at the NTCIR SciClaimEval
2026-08-30 2608.29661v1 OmniClimate-TC: Physics-Aware Visual Abstractions for Multimedia Reasoning over Tropical Cyclones
2026-08-29 2608.29088v2 HANIA: Planner-Guided Multimodal Graph Evidence Selection for Grounded Question Answering
2026-08-28 2608.27867v1 CoRe-MoE: Compact Reusable MoE for Continual Multimodal Instruction Tuning
2026-08-27 2609.25040v1 BananaVLM: A Domain-Adapted Vision Language Model for Banana Crop Disease Diagnosis
2026-08-27 2608.26820v1 LLaVAFlow: Preserving Latent Alignment Flow for Parameter-Efficient Multimodal Fine-Tuning
2026-08-25 2608.25198v1 Tunable Tool-Call Rates in LLM Agents via Representation Steering
2026-08-16 2608.15516v1 UniFed-VLM: Federated Instruction Tuning for Vision-Language Models with Multiple Heterogeneity
2026-08-14 2608.14198v1 MINT: A Universal Zero-Shot Predictor for Transaction Data
2026-08-10 2608.09907v1 DistMoE: Private-data Rehearsal-free Routing in Mixture-of-Experts for Distributed Instruction Tuning
2026-08-08 2608.08107v1 NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs
2026-08-06 2608.06246v1 A Six-Dimensional Taxonomy of Post-Training Adaptation Techniques with Applications in AI Governance
2026-08-06 2608.05691v1 SciQNet: Two-Stage Multimodal Adaptation for Scientific Image Quality Assessment
2026-08-03 2608.02790v1 Confident but Unreliable: A Behavioral Safety Audit of Vision-Language Models on Brain MRI
2026-08-03 2608.01635v1 Mitigating Visual Degradation in MLLMs via Spatial-Spectral Visual Anchor Learning
2026-08-02 2608.01437v1 Beyond Routing Saturation: A Long-Horizon Class-Incremental Perspective on Expert Routing in Multimodal Contin
2026-07-29 2607.26947v2 Progressive Multimodal Alignment for Continual Instruction Tuning
2026-07-29 2607.26596v1 Decoupled Visual Processing: Efficient Multimodal Adaptation via Modality-Specific Transformer Substitution
2026-07-29 2608.00068v1 SafeBuild-Bench: A Temporal-Robust Construction Safety Benchmark with Graph-Enhanced Data Mining
2026-07-28 2607.25947v1 A Cost-Effective Multimodal LLM Reasoning Framework for Question Answering over Irregular Clinical Time Series
2026-07-27 2607.23917v1 Gaze-to-text Generation: Beyond Categorical Decoding of Human Attention
2026-07-15 2607.13860v1 Towards Enhancing 3D Spatial Reasoning in Medical Multimodal Large Language Models
2026-07-13 2607.11102v1 CHARM: Charge Calibration and Acoustic Rescue for LLM-based Multimodal Sarcasm Detection
2026-07-11 2607.10308v1 Generalize LMMs to Versatile Visual Modalities via Fabricated Modality Synthesis
2026-07-11 2607.10299v1 Empowering Long-form Omni-modal Understanding with Robust Audio Perception

abs:"parameter-efficient" AND abs:"multimodal" — 命中 38，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-04 2610.05129v1 Representation--Behavior Alignment for Explainable Weakly-Supervised Video Anomaly Detection
2026-09-27 2609.33640v1 SafeMol: Dual-Modality Safety Alignment for Molecular Multimodal Models
2026-09-12 2609.13654v1 Multimodal Foundation Models Adaptation based on Domain-Aware Relaxed Orthogonal Subspace for Remote Sensing
2026-09-06 2609.06497v1 Vision-Guided Text Prompt Tuning for Multimodal Sentiment Analysis
2026-09-05 2609.06188v1 MVFA: A Multi-View Text-Guided Multimodal Fusion LLM Adapter for Sentiment Analysis and Emotion Recognition
2026-09-02 2609.05539v1 Dual-Latent Memory Routing for Vision-Language Reasoning
2026-09-02 2609.02101v1 Federated LoRA Adaptation of BiomedCLIP Across Four International Chest X-Ray Cohorts
2026-08-31 2608.30712v1 GUIDE: Guiding Internal Evidence with Language Instructions
2026-08-30 2608.29802v1 Foundation and Multimodal Large Language Models for Face Presentation and Morph Attack Detection
2026-08-29 2608.29357v1 LiteSearch-VL: Small Multimodal Search Agents via Trajectory Distillation and Synthetic Step-DPO
2026-08-28 2608.27867v1 CoRe-MoE: Compact Reusable MoE for Continual Multimodal Instruction Tuning
2026-08-19 2608.19355v1 GRACE: Grounded Reasoning via Adapter Composition and Evidence-Aware Calibration for Educational Visual Questi
2026-08-18 2608.18386v1 TTSD-FAR: Test-Time Self-Distillation with Fisher-Anchored Restoration for Missing-Modality Emotion Recognitio
2026-08-18 2608.17514v2 SE-MoLoRA: Shared-Expert LoRA Adapters for Domain-Specific Photographic Assessment
2026-08-16 2608.15516v1 UniFed-VLM: Federated Instruction Tuning for Vision-Language Models with Multiple Heterogeneity
2026-08-10 2609.26182v1 Modality-Gated Deep Adapters: Adding a Modality to a Frozen Embedding Model with Exact Preservation
2026-08-08 2608.08244v2 FemWear: A Parameter-Efficient Wearable Foundation Model for Women's Health
2026-08-06 2608.06246v1 A Six-Dimensional Taxonomy of Post-Training Adaptation Techniques with Applications in AI Governance
2026-08-04 2608.04130v1 Radar4D-VLM: Proposal-Grounded Temporal 4D Radar Reasoning Across Frozen Language Models
2026-08-01 2608.00847v1 Models as Tools: An Agentic Coordination Framework for Unified Multimodal Visual Tracking
2026-07-29 2607.26596v1 Decoupled Visual Processing: Efficient Multimodal Adaptation via Modality-Specific Transformer Substitution
2026-07-25 2607.23245v1 FedTaste: Topology-Aware Structural Transfer for Multimodal Federated Learning with Missing Modalities
2026-07-21 2607.19270v1 GUIDED Network-Agnostic Feature Initialization for Spatial Transferability in GNN-based Models
2026-07-21 2607.18716v1 Continual Video-MLLM Adaptation over Evolving Domains
2026-07-16 2607.15241v1 Beyond the Leaderboard: Design Lessons for Trustworthy Multimodal VQA
2026-07-15 2607.13466v1 PQFA: Parallel Quantum Feature Augmentation of Fused Representations for Multimodal Classification
2026-07-13 2607.11839v1 LoRA-Based Cascaded Multimodal Fusion for Action Recognition in Medical Training Environments
2026-07-10 2607.09583v1 Promptable Concept Segmentation from Above: Evaluating SAM 3's Zero-Shot and One-Shot Capabilities in Remote S
2026-07-07 2607.05736v1 Multimodal Molecular Representation Learning with Graph Neural Networks, Deep & Cross Networks, and SMILES Emb
2026-07-03 2607.02930v1 CL-Anomaly: Layer-Adaptive Mixture-of-Experts with Multimodal Large Language Model for Continual Learning in A

abs:"catastrophic forgetting" AND abs:"multimodal" — 命中 21，其中新增 21

日期 arXiv 标题
2026-09-29 2609.37889v1 ReCAP: Retrieval-Guided Capability Reuse for Multimodal Continual Instruction Tuning
2026-09-22 2609.25841v1 Metric-Bench: Exploring In-context Spatial Metric Reasoning in VLMs for Indoor Scenes
2026-09-21 2609.24813v1 INTCORT: Training-Free Spatial Reasoning Enhancement for Vision-Language Models via Input Transformations and
2026-09-07 2609.07009v1 NeuCME: Toward Dynamic Multimodal Continual Learning via Neural Combinatorics of Multiple Experts
2026-08-27 2608.26820v1 LLaVAFlow: Preserving Latent Alignment Flow for Parameter-Efficient Multimodal Fine-Tuning
2026-08-26 2608.26095v1 A Visual Dependence-Aware Framework for Multimodal Unsupervised Continual Post-Training
2026-08-24 2608.23242v1 Mind the Couch! Eliciting MLLM Reasoning in Interior Design via Weak-to-Strong Task Vector Injection
2026-08-12 2608.11758v1 AWARe: Mitigating Catastrophic Forgetting via Activation-Weighted Adaptive REtention
2026-08-04 2608.03660v1 Taming the Implicit: Dual-Channel Risk-Aware Reinforcement Fine-Tuning for Continual Multimodal Post-Training
2026-07-30 2607.28050v1 IndustryForge-27B: A Domain-Enhanced Multimodal Foundation Model for Industrial CAD
2026-07-30 2607.27665v1 FedOGL: Combating Catastrophic Forgetting in Federated Open-World Multimodal Graph Learning
2026-07-21 2607.18716v1 Continual Video-MLLM Adaptation over Evolving Domains
2026-07-13 2607.12112v2 Continual Learning with Elastic Regularization and Synthetic Replay for Federated MLLM Fine-Tuning
2026-07-05 2607.04364v3 RL Forgets! Towards Continual Policy Optimization
2026-07-05 2607.20511v1 SiGMA: Sign-Guided Merging and Adaptation for Multimodal Continual Instruction Tuning
2026-07-04 2608.26155v1 VFA: Empowering Multilingual MLLMs via Vision-Free Adaptation
2026-07-01 2607.00434v1 Information-Regularized Attention for Visual-Centric Reasoning
2026-07-01 2607.00302v1 Wake up for Touch! Mask-isolated Tactile Alignment Learning in MLLMs
2026-07-01 2607.00293v1 Rosetta: Composable Native Multimodal Pretraining
2026-06-23 2606.24963v1 Curvature-Guided Mixing for MLLM Adaptation
2026-06-16 2606.18472v1 Domain Generalizable Adaptation of 3D Vision-Language Models via Regularized Fine-Tuning

abs:"preference alignment" AND abs:"vision-language" — 命中 8，其中新增 8

日期 arXiv 标题
2026-09-23 2609.31746v1 VisionPsy-Nano: Improving Accuracy, Efficiency, and Reliability in On-Device Vision-Language Models
2026-09-08 2609.08188v1 Bridging the Semantic-Utility Gap in Multimodal RAG via Generator-in-the-Loop Alignment
2026-09-06 2609.13259v1 TryOnReward: Learning Foveated Consistency for Reinforcement Fine-Tuning of Virtual Try-On
2026-09-04 2609.04961v1 SAM-D2Q: Aligning Multimodal Doc2Query with Search Demand and Conversion for E-commerce
2026-08-26 2608.25452v1 VGA-BenchV2: An Expanded Unified Benchmark and Multi-Model Framework for Evaluating Video Aesthetics and Gener
2026-08-14 2608.14260v1 Personalized Digital Semantic Communication for Image Transmission with Vision-Language Models
2026-07-02 2607.01721v1 CoRe: Combined Rewards with Vision-Language Model Feedback for Preference-Aligned Reinforcement Learning
2026-06-24 2606.28401v1 Vision-driven Preference Synthesis for Mitigating Hallucinations in VLMs

07 · 通用评测与幻觉

abs:"hallucination" AND abs:"multimodal large language" — 命中 78，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-04 2610.05610v2 SEA-LM: Egocentric Spatial Audio Understanding for Wearable Microphone Arrays
2026-10-03 2610.06972v1 BoT-Feedback: Grounding Multimodal Reasoning in Biomechanical Evidence for Explainable Human Action Feedback
2026-10-02 2610.02887v1 Revealing Epistemic Uncertainty in MLLMs via Causal-Invariant Masking
2026-09-29 2609.36562v1 ThinkingGuard: Decoding Implicit Hazards via Step-by-Step Risk Attribution in Multimodal Large Language Models
2026-09-28 2609.35536v1 Look Before You Judge: Training-Free Region Mining for Grounded and Explainable Deepfake Detection
2026-09-28 2609.34330v2 MiCo: Mutual Information Coverage Optimization through Semantic Erasure Modeling for Efficient MLLM Inference
2026-09-24 2609.29940v1 Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration
2026-09-23 2609.28570v1 DEEPO: Dual-Entropy Enhanced Policy Optimization for Hallucination in MLLMs
2026-09-18 2609.21828v1 Touvigation: Embodied Adaptive Object Acquisition for Blind and Low-Vision Users in Unfamiliar Indoor Environm
2026-09-18 2609.21675v1 DRT: Dense Reasoning Trace for Efficient and Grounded Multimodal Reasoning
2026-09-15 2609.17269v1 Semantic-Spatial Agreement Verification for Mitigating Object Hallucination in Multimodal Large Language Model
2026-09-15 2609.17248v1 Video-HolmesV2: Can MLLMs Reason with Spatio-Temporal Audio-Visual Evidence in Long Videos?
2026-09-15 2609.16601v1 SAVOR: Self-Aware Visual Grounding via Confidence-Calibrated Reinforcement Learning for Multimodal Hallucinati
2026-09-12 2609.13849v1 UniCAR-RL: Seeing Better before Thinking Deeper in Visual Mathematics
2026-09-10 2609.11244v1 OmniHallu: Unified Hallucination Detection for Cross-Modal Comprehension and Generation in Multimodal Large La
2026-09-10 2609.11154v1 Multi-Faceted Evaluation and Mitigation of Emotion Hallucinations in MLLMs
2026-09-08 2609.08300v1 Human-Centric Image Captioning with Subject-Centered Spatial Understanding
2026-09-05 2609.09206v2 MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads
2026-09-03 2609.03414v1 StrixAE: An Intelligent Agent for Audio Enhancement under Complex Distortion Coupling in Real-World Scenarios
2026-08-31 2609.00231v1 Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM
2026-08-31 2608.30653v1 Fine-Grained Multi Image Object Hallucination Benchmark
2026-08-28 2609.29607v1 STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models
2026-08-27 2608.28707v1 ReVA: A Region-Aware Visual Assistant for Visually Grounded Question Answering
2026-08-26 2608.25986v1 Multi-Granularity Context-Enhanced RAG over Multimodal Knowledge Graphs
2026-08-25 2608.23974v1 Boot-and-Feedback Framework for Generalist-Expert Model Collaboration in Breast Ultrasound Diagnosis
2026-08-24 2608.23242v1 Mind the Couch! Eliciting MLLM Reasoning in Interior Design via Weak-to-Strong Task Vector Injection
2026-08-23 2608.22367v1 Context-Aware Cluster Decoding: Semantic Anchor-Driven Coherence in dMLLMs
2026-08-21 2609.27890v1 RelCheck: Dual-Evidence Spatial Grounding for VLM Hallucination Correction
2026-08-19 2608.18586v1 OmniHandwritingOCR: A Diagnostic Benchmark for Evaluating Multimodal LLMs in Handwritten OCR Scenarios
2026-08-18 2608.17895v1 BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models

abs:"MMMU" — 命中 15，其中新增 15

日期 arXiv 标题
2026-10-05 2610.05897v1 Fitting Vision Adapters at Frontier Scales
2026-09-29 2609.38413v1 VidHarness: Evolving Agent Harnesses for Cost-Efficient Long Video Understanding
2026-09-29 2609.36651v1 FocusVTC: Efficient and High-Performance Visual Text Compression with Adaptive Resolution
2026-09-27 2609.33414v1 TTRSD: Test-Time Reinforcement Learning with Self-Distillation for Vision-Language Models
2026-09-22 2609.26333v1 Disaggregated Quantization: Specializing LLM Prefill and Decode
2026-09-14 2609.15051v1 Not All Prompts Are Equal: Exploration-Guided Prompt Scaffolding for Multimodal Reinforcement Post-Training
2026-09-13 2609.14258v1 SPARK: Representation-Level KV Memory Alignment for Safer Vision-Language Models
2026-08-27 2608.26495v2 Video-FLAIR: Not Whether to Reason, But How
2026-08-24 2608.23268v1 Dual-Grained Agent Memory and Shapley Context Attribution for Multimodal Agentic Learner
2026-08-02 2608.01207v2 It's the Decoding Format, Not the Perturbation: Auditing Consistency-Based Selection for Vision-Language Test-
2026-07-15 2607.14333v1 SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning
2026-07-09 2607.08215v1 On the Limitations of Non-GPU AI Accelerators for Large-Model Inference: A Field Study of MoE and Multimodal S
2026-06-25 2606.27376v1 Ask, Solve, Generate: Self-Evolving Unified Multimodal Understanding and Generation via Self-Consistency Rewar
2026-06-16 2606.18385v1 CaVe-VLM-CoT: An Interpretable Vision-Language Model Framework
2026-06-12 2606.14629v3 When Good Verifiers Go Bad: Silent Negative Transfer in Verifier-Guided VLM Training

abs:"MMBench" — 命中 9，其中新增 9

日期 arXiv 标题
2026-10-07 2610.09361v1 From Chunks to Functional Evidence: Function-Aware Retrieval for EDA Documentation QA
2026-09-30 2609.40362v1 Multimodal Flow: Unified Flow Modeling of Language and Vision in Embedding Spaces
2026-09-15 2609.16601v1 SAVOR: Self-Aware Visual Grounding via Confidence-Calibrated Reinforcement Learning for Multimodal Hallucinati
2026-08-27 2608.28707v1 ReVA: A Region-Aware Visual Assistant for Visually Grounded Question Answering
2026-08-03 2608.01979v2 ET-Prune: Evidence-Aware Dynamic Budgeting for Visual Token Pruning in Text-Rich MLLMs
2026-07-27 2607.24424v1 MAViE: A Multi-scale Adaptive Vision Encoder for Fine-grained Visual Perception and Efficient Multimodal Reaso
2026-07-15 2607.14333v1 SD-MAR: Multi-image Analytical Reasoning via Synthetic Data and Reinforcement Learning
2026-07-02 2607.01813v1 MMBench-Live: A Continuously Evolving Benchmark for Multimodal Models
2026-06-29 2606.30084v1 One Forward Beats Two: InnerZoom for Accurate and Efficient GUI Grounding

abs:"POPE" — 命中 28，其中新增 28

日期 arXiv 标题
2026-09-30 2609.40362v1 Multimodal Flow: Unified Flow Modeling of Language and Vision in Embedding Spaces
2026-09-29 2609.38485v1 Beyond Layers: Position-Resolved Gradient Conflict and Position-Aware Modulation for Unified Multimodal Models
2026-09-15 2609.17269v1 Semantic-Spatial Agreement Verification for Mitigating Object Hallucination in Multimodal Large Language Model
2026-09-15 2609.16601v1 SAVOR: Self-Aware Visual Grounding via Confidence-Calibrated Reinforcement Learning for Multimodal Hallucinati
2026-09-10 2609.11910v2 From Protocols to Evidence: Bounded Claims for AI in Service of the Common Good
2026-09-01 2610.00024v1 Encoded but Disconnected: Decomposing Vision-Language Model Failures under a Patching Null
2026-08-31 2609.00231v1 Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM
2026-08-30 2608.29996v1 Partition-Aware Unlearning for Removing Spurious Correlations in Large Vision-Language Models
2026-08-30 2608.29924v1 Hallucination Mitigation for Large Vision-Language Models via Implicit Feature Stabilization
2026-08-29 2608.29193v1 HalluPrism: When Multimodal Uncertainty Should Diagnose, Not Decide
2026-08-29 2608.29092v1 EviAnchor: Mitigating Hallucinations in Large Vision-Language Models via Regional Visual Evidence Compensation
2026-08-27 2608.28707v1 ReVA: A Region-Aware Visual Assistant for Visually Grounded Question Answering
2026-08-26 2608.25865v1 Instability of Böhm's Einstein metrics
2026-08-08 2608.08167v1 Wiener Representation Filtering for VLM Hallucination Suppression
2026-07-29 2607.27145v1 Explainable and Resource-Efficient Spatial Reasoning in Multimodal LLMs for Decision-Critical Applications
2026-07-29 2607.26596v1 Decoupled Visual Processing: Efficient Multimodal Adaptation via Modality-Specific Transformer Substitution
2026-07-28 2607.25527v1 Argus-Unified: Towards A Compact and Economical Unified Model for Image Understanding and Generation
2026-07-18 2607.16727v1 Constraint-Anchored Reasoning Traces
2026-07-14 2607.12962v1 Form, Not Content? A Preregistered, Placebo-Controlled Evaluation of Learned Error-Conditioned Self-Repair Thr
2026-07-13 2607.11791v1 Cosmology: 100 years after A. A. Friedmann
2026-07-09 2607.08059v1 When Thinking Hurts: Epistemic Signals in the Reasoning Chains of Visual Language Models
2026-07-07 2607.06726v1 A Good Initialization is All You Need for Faithful Visual Attribution
2026-07-05 2607.04163v1 SeeMe: Mitigating Hallucinations in Large Vision-Language Models through Effective Visual Token Engineering
2026-07-03 2607.03143v1 Text as Partial Constraint: Core-Residual Alignment for Robust Vision-Language Learning
2026-07-01 2607.00784v1 LeVLJEPA: End-to-End Vision-Language Pretraining Without Negatives
2026-06-28 2606.29431v5 FADE: Mitigating Hallucinations by Reducing Language-Prior Dominance in Large Vision-Language Models
2026-06-27 2606.28862v2 HKVLM: Faithful Query--Region Binding for Frozen-Detector Visual Grounding
2026-06-14 2606.15821v2 The Truth Stays in the Family: Enhancing Contextual Grounding via Inherited Truthful Heads in Model Lineages

abs:"visual question answering" AND abs:"hallucination" — 命中 18，其中新增 18

日期 arXiv 标题
2026-09-29 2609.37685v2 PAIQ: Patch-Aligned Semantic Injection via Residual Rotation
2026-09-16 2609.18562v1 Sim-to-Real Traffic Scene Understanding by Decoupling Semantics from Caption Generation with V-JEPA
2026-09-15 2609.16601v1 SAVOR: Self-Aware Visual Grounding via Confidence-Calibrated Reinforcement Learning for Multimodal Hallucinati
2026-08-31 2608.30475v1 ImageEval 2026: Culturally Grounded Arabic Multimodal Evaluation
2026-08-28 2609.29604v1 PROVE: Proof-guided Regime-aware Operator Verification for Hallucination Detection in Medical Visual Question
2026-08-27 2608.28707v1 ReVA: A Region-Aware Visual Assistant for Visually Grounded Question Answering
2026-08-18 2608.17427v1 Counterfactual Anatomy-guided Spatial-Temporal Decoding for Annotation-Free Hallucination Mitigation in Medica
2026-08-17 2608.16805v1 Diagnosing Dense Same-Class Attribute Misbinding in Large Vision-Language Models
2026-08-11 2608.10964v1 CARE: Confidence-Aware Reasoning for Reliable Medical VQA
2026-07-30 2607.28374v1 LEDGERMIND: Provenance-Constrained Multimodal Agentic Reasoning with a Structured Evidence Ledger
2026-07-29 2607.26885v1 SCALPEL: Semantic Cross-modal Alignment via LLM-Powered Encoder Learning for Medical Vision-Language Represent
2026-07-14 2607.12319v1 DM-KG: A Novel Method for Boosting Spatial Cognition of Vision-Language Models in Street View Imagery
2026-07-05 2607.04163v1 SeeMe: Mitigating Hallucinations in Large Vision-Language Models through Effective Visual Token Engineering
2026-06-30 2606.32012v1 CoMet: Context and Multiplicity Decomposition for Multimodal Uncertainty Estimation
2026-06-25 2606.27373v1 Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models
2026-06-23 2606.24115v1 A Benchmark for Hallucination Detection in VLMs for Gastrointestinal Endoscopy
2026-06-22 2606.23354v1 Faithful Grounded Visual Reasoning via Learned Proxy-Tokens
2026-06-17 2606.19584v1 Language-Instructed Vision Embeddings for Controllable and Generalizable Perception

abs:"vision-language model" AND abs:"evaluation" AND abs:"bias" — 命中 39，其中新增 29（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-04 2610.05023v1 Look Where You Say You're Looking: Self-Grounded Attention for Visual Reasoning
2026-09-28 2609.36101v1 One Geometry, Different Outcomes: Readout-Dependent Effects of the Modality Gap in Vision-Language Models
2026-09-23 2609.27682v2 Gender Bias in Vision-Language In-Context Learning
2026-09-23 2609.31751v1 EgoTSR++: Egocentric Spatiotemporal Reasoning for Task Progress Understanding
2026-09-21 2609.24485v2 VPRune: Efficient Training-free Pre-LLM Visual Token Pruning
2026-09-15 2609.16651v1 Mechanism-Level Evaluation for Vision-Language Models: Controlled Activation-Replacement Diagnosis of Gender B
2026-09-15 2609.16647v1 ViD: Vision-Dominant Gender Bias Mitigation for Large Vision-Language Models
2026-09-04 2609.05761v1 GeoContext: One Context Ladder, Two Failure Modes in Vision-Language Geolocation: Flat Reliance on User-Provid
2026-09-01 2609.01691v1 FairLens: Benchmarking Fairness in Vision-Language Models for High-Stakes Decision-Making
2026-09-01 2609.01318v1 Reliability Challenges in Diffusion Vision-Language Models
2026-08-30 2608.29590v1 Guardrail-Agnostic Societal Bias Evaluation in Large Vision-Language Models
2026-08-26 2608.25375v1 GGSS: Geodesic-Gated Spherical Steering for Inference-Time Debiasing of Generative Vision-Language Models
2026-08-17 2608.16081v2 SafeGesture: Evaluating Fine-Grained Hand Gesture Understanding in Vision-Language Models through Scenario-Con
2026-08-14 2608.13969v1 PPOM: Marginalizing Patch-Grid Phase for CLIP-Based Generalizable Vision-Language Prompt Tuning
2026-08-13 2608.13267v1 How Do VLMs Behave When Blind or Misled? Behavioral Evaluation of VLMs on Scientific Figures
2026-08-12 2608.21415v1 Mitigating Bias in Large Vision-Language Models via Counterfactual Ensemble Decoding
2026-08-12 2609.26210v1 Same Chart, Different Story: Bias in Vision-Language Chart Interpretation
2026-08-11 2608.11074v1 CapProbe: Evaluating Detailed Image Captions via Full-Scene Dense Question Answering
2026-08-05 2608.04510v1 GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs
2026-08-04 2608.03875v1 Enhancing VLM Reward Models Through Structure-Aware Fine-Tuning
2026-08-03 2608.01661v1 FairForensics: Seeing Expressions and Parsing Demographics via Vision-Language Modeling for Generalizable Fair
2026-07-31 2608.00119v1 Counting the Cost of War Under Satellite Embargo: Zero-Shot Estimation of Impacted Infrastructure
2026-07-30 2607.28609v2 OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models
2026-07-30 2607.28211v1 Scaling Vision-Language Models Is Not Enough to Mitigate Bias
2026-07-29 2609.17572v1 Disentangling Algorithmic Bias from Archival Artifacts: A Controlled Audit of Vision-Language Model Valuation
2026-07-21 2607.18673v3 MissingBench-Verified: Probing Vision-Language Models' Inability to Detect Missing Object Parts
2026-07-13 2607.11228v1 DeepBias: Adaptive In-depth Probing of Social Biases in LVLMs
2026-07-13 2607.11044v2 RetroHolmes: When Semantic Plausibility Fails Retrospective Physical Process Reasoning
2026-07-04 2607.03831v1 How Do Diffusion Classifiers Decide? A Bias-Centric Evaluation

uav · 无人机主线

abs:"UAV" AND abs:"vision-language" — 命中 44，其中新增 16（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-01 2610.02451v1 A Simulation-Grounded Agentic VLM Framework for Wildfire Monitoring and Reporting
2026-09-24 2609.30629v1 FRESHLATENT: Channel-Aware Latent Adaptation for Resource-Constrained Embodied VLM Perception
2026-09-16 2609.18448v1 From Pixels to Semantics: Edge AI for UAV-Based Critical Infrastructure Inspection
2026-09-16 2609.18326v1 UAVs Meet Embodied Intelligence: Bridging Human Intents and Flying Dynamics Via Harnessing Physical-Digital AI
2026-09-16 2609.18210v2 Understanding Dynamic Scenes at Gigapixel Scale: Wide-Area Spatio-Temporal Perception from UAVs
2026-09-08 2609.08402v1 Towards Embodied Air-Ground Cooperative Object Search: Benchmark, Dataset and Agentic Method
2026-09-08 2609.08164v1 Dual-Layer Semantic-Spatial Belief Mapping for Aerial Object Goal Navigation
2026-08-24 2608.22678v1 RACO: Reliability-Aware Coarse-Goal Optimization for Inspection-Oriented UAV Vision-Language Navigation
2026-08-13 2608.12835v1 AirForesight: Current-to-Future Spatial Map Imagination with Cross-Space Planning Consistency for UAV-VLN
2026-08-12 2608.14721v1 AeroGround: A Comprehensive Benchmark for Aerial-Ground Collaborative Reasoning
2026-08-10 2608.09564v1 From Semantic Grounding to Decision Optimization: A Unified Framework for Long-Horizon UAV Vision-Language Nav
2026-08-08 2608.08045v1 Lingjing: A Simulation Testbed for Multi-Agent Embodied Tasks in Open-Ended Cities
2026-08-05 2608.04825v1 Deliberate Before You Fly: Vision-Guided Spatial Deliberation for UAV See-and-Reach Navigation
2026-08-03 2608.01906v1 Assessing the Benefits of Combining Advanced Deep Learning Techniques for Post-Disaster Building Damage Assess
2026-08-03 2608.01802v1 CoNav-UAV: Cooperative Dual-Altitude Aerial Navigation via Stackelberg Learning
2026-07-30 2607.27597v1 A Systems Engineering Framework for Vision-Language-Enabled UAV Triage and Disaster Response

abs:"aerial" AND abs:"vision-language model" — 命中 30，其中新增 20

日期 arXiv 标题
2026-10-04 2610.05342v1 IRSTD-Agent: Agentic Infrared Small Target Detection via Zoom-Guided Interaction Learning
2026-09-20 2609.23655v1 Reassessing Global Gradient-Norm Imbalance in BLIP Fine-Tuning Across Physical Domains
2026-09-16 2609.18448v1 From Pixels to Semantics: Edge AI for UAV-Based Critical Infrastructure Inspection
2026-09-16 2609.18210v2 Understanding Dynamic Scenes at Gigapixel Scale: Wide-Area Spatio-Temporal Perception from UAVs
2026-09-10 2609.11310v1 Your Model Already Knows Don't Teach It, Learn to Ask It: Soft Prompting for Few-Shot Adaptation of Vision-Lan
2026-09-08 2609.08402v1 Towards Embodied Air-Ground Cooperative Object Search: Benchmark, Dataset and Agentic Method
2026-09-08 2609.08164v1 Dual-Layer Semantic-Spatial Belief Mapping for Aerial Object Goal Navigation
2026-09-01 2609.00628v1 Restrict, Don't Retrain: Inference-Time VLM Guidance for Zero-Shot Aerial Segmentation
2026-08-12 2608.14721v1 AeroGround: A Comprehensive Benchmark for Aerial-Ground Collaborative Reasoning
2026-08-04 2608.04175v1 TriCLE: Tri-Modal Vision-Language Reasoning for Edge-Deployed Fine-Grained Clustering
2026-08-03 2608.02039v2 RSVideo: Are Your Vision-Language Models Ready for Remote Sensing Videos?
2026-08-03 2608.01906v1 Assessing the Benefits of Combining Advanced Deep Learning Techniques for Post-Disaster Building Damage Assess
2026-07-21 2607.19288v1 No Training, Better Flights: Test-Time Scaled VLMs for UAV Navigation
2026-07-13 2607.12177v1 The Emerging Paradigm of Geospatial Foundation Models: From Pre-Training to Agentic Reasoning
2026-07-09 2607.08359v1 FSD-VLN: Fast-Slow Dual-System Modeling for Aerial Long-Horizon Vision-Language Navigation
2026-07-07 2607.07350v1 Towards Reliable Aerial Ground Vehicle Collaboration: An Integrated Planning and Autonomy Framework for Field
2026-07-02 2607.02718v1 Diagnosing Aerial-View Object Detectors with Foundational Image Generative Models
2026-07-01 2607.00338v1 DroneFINE: Domain-Aware Parameter-Efficient Fine-Tuning of Vision-Language Detectors for Drone Images
2026-06-19 2607.14125v1 CARPRT: Class-Aware Zero-Shot Prompt Reweighting for Black-Box Vision-Language Models
2026-06-17 2606.19277v1 A Unified Framework for Efficient Remote Sensing Visual Question Answering: Adapting Dual, Hybrid, and Encoder

abs:"drone" AND abs:"vision-language" — 命中 9，其中新增 7

日期 arXiv 标题
2026-09-25 2609.31005v1 TRACKGRAPH: Online Open-Vocabulary 3D Scene Graphs via Image-Space Tracking
2026-09-16 2609.18139v1 Multi-View Mixture-of-Experts with Vision-Language Reranking for Cross-View Object Geo-Localization
2026-09-07 2609.07780v2 DroneGround: Open-Vocabulary Drone Payload Characterization Using Synthetic Data and Grounded Vision-Language
2026-08-27 2608.26722v1 UniGeo: A Multi-modal Large Language Model for Text-Guided Cross-View Geo-Localization
2026-08-10 2608.09270v1 GRASP: Granularity-Aware Region Alignment and Semantic Prototype Learning for Fine-Grained Cross-Modal Underst
2026-07-07 2607.06706v1 Vision Language Action (VLA) Models for Unmanned Aerial Robotics and Bimanual Manipulation: A Review
2026-07-01 2607.00338v1 DroneFINE: Domain-Aware Parameter-Efficient Fine-Tuning of Vision-Language Detectors for Drone Images

abs:"remote sensing" AND abs:"vision-language model" — 命中 33，其中新增 30（只取了最近 30 条，新增数是下界）

日期 arXiv 标题
2026-10-06 2610.09085v1 Anaximander: Interactively Running Geospatial Deep Learning Models on Any Compute Backend
2026-10-05 2610.06196v1 EORestore-Agent: Fidelity-Guided Agentic Restoration of Remote Sensing Images with Composite Degradations
2026-09-29 2609.38603v1 Aperture: Training-Free Multiscale Concept Bottlenecks for Remote Sensing
2026-09-28 2610.00302v1 Decoding the Disaster: Multi-Task Geospatial Reasoning with Vision-Language Models and Crowdsourced Imagery fo
2026-09-26 2609.32813v1 USAI-Quant: A Quantitative Reasoning Benchmark for Vision-Language Models in Built Environments
2026-09-23 2609.28230v2 A Unified Framework and Dataset for Oriented Object Visual Grounding in Remote Sensing
2026-09-16 2609.18210v2 Understanding Dynamic Scenes at Gigapixel Scale: Wide-Area Spatio-Temporal Perception from UAVs
2026-09-15 2609.16486v1 VPRef: A Cross-Domain Benchmark for Referring Remote Sensing Image Segmentation
2026-09-13 2609.14523v1 Selective Tool Use for Agentic Change Visual Question Answering in Remote Sensing
2026-09-09 2609.09876v1 From Pixels to Hierarchical Sequences: Quadtree Mask Encoding for Vision-Language Binary Change Detection
2026-09-03 2609.03391v2 Exploring the Potential of Contrastive Language-Image Pre-training for Multi-Source Remote Sensing Data
2026-09-01 2609.01289v1 Agentic Multimodal Models for Environmental Hyperspectral Unmixing
2026-08-26 2608.25485v1 Semi-Supervised Adaptation of Vision-Language Models for Image Classification
2026-08-13 2608.13344v1 LongEarth-R1: Benchmarking and Aligning Vision-Language Models for Long-Horizon Earth Observation Reasoning
2026-08-07 2608.06959v1 Summarize First, Download Later: Onboard VLMs for Bandwidth-Efficient Earth Observation
2026-08-05 2608.04791v1 On the Effectiveness of Adaptation Strategies for VLM-Based Federated Learning in Remote Sensing
2026-08-04 2608.03911v1 UniEvo-RS: Omni-Prompt Unified Remote Sensing Segmentation with Representative Exemplar-Driven Prototype Evolu
2026-08-03 2608.02039v2 RSVideo: Are Your Vision-Language Models Ready for Remote Sensing Videos?
2026-07-31 2607.29192v1 Locally Consistent Transductive Information Maximization for Few-Shot Remote Sensing Scene Classification
2026-07-23 2607.21036v1 GeoThreat: Transferable Targeted Adversarial Attacks on Large Vision-Language Models for Remote Sensing Image
2026-07-18 2607.16819v1 FUSAR-R1: A Large-Scale Reasoning Model for Intelligent Interpretation of SAR Images
2026-07-17 2607.15942v2 More with Less: a Large Scale Remote Sensing VLM with a Simple Recipe
2026-07-13 2607.24810v1 RRS-10K: A Multitask Vision-Language Model Benchmark for Rare Remote Sensing Image Interpretation
2026-07-11 2607.10120v1 WeaveEarth: Structured Evidence Construction and Reasoning for Training-Free UHR Remote Sensing Understanding
2026-07-07 2607.06485v1 AirflowAttack: Thermal-Airflow Adversarial Perturbations against Infrared Remote-Sensing Vision-Language Model
2026-07-02 2607.02718v1 Diagnosing Aerial-View Object Detectors with Foundational Image Generative Models
2026-06-30 2606.31976v1 TreeAgent: A Generalizable Multi-Agent Framework for Automated Bias Labeling in Forestry via Compiled Expert R
2026-06-26 2606.28266v1 RSICCLLM: A Multimodal Large Language Model for Remote Sensing Image Change Captioning
2026-06-17 2606.19277v1 A Unified Framework for Efficient Remote Sensing Visual Question Answering: Adapting Dual, Hybrid, and Encoder
2026-06-15 2606.17246v1 GeoDisaster: Benchmarking Orchestrated Agents for Operational Disaster Geo-Intelligence

abs:"UAV" AND abs:"multimodal large language model" — 命中 10，其中新增 6

日期 arXiv 标题
2026-08-14 2608.13980v1 FIRM: Fine-Grained Intra-Token Representation of Masks for Remote Sensing Reasoning Segmentation
2026-07-22 2607.19857v1 Memory-Augmented Multimodal Large Language Models for Small Object Understanding in Streaming Aerial Videos
2026-07-19 2607.17386v1 SkyVLaM: Multimodal Large Language Model for UAV Video Understanding in Remote Sensing
2026-07-14 2607.12477v2 Self in Space: Benchmarking Self-Awareness and Spatial Cognition in UAV Embodied Intelligence
2026-07-01 2607.00416v1 DroneIQA-VLE: Multi-Task Drone Image Quality Assessment via Vision-Language Ensemble
2026-06-26 2606.28049v1 AirGroundBench: Probing Spatial Intelligence in Multimodal Large Models under Heterogeneous Multi-View Embodie

abs:"aerial" AND abs:"visual question answering" — 命中 3，其中新增 2

日期 arXiv 标题
2026-09-16 2609.18210v2 Understanding Dynamic Scenes at Gigapixel Scale: Wide-Area Spatio-Temporal Perception from UAVs
2026-06-17 2606.19277v1 A Unified Framework for Efficient Remote Sensing Visual Question Answering: Adapting Dual, Hybrid, and Encoder

本文件由 tools/watch.py 生成，重跑即可刷新。收录进正文前，ID 与标题一律以 arXiv API 为准。