ICML 2026 accepted papers on Vision-language and multimodal
549 of the 6,341 papers accepted at ICML 2026 (8.7%) are on Vision-language and multimodal, compared with 5.3% at ICML 2025. They fall into 6 topics.
See ICML 2026 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Multimodal LLMs · 266 papers
- BabyVision: Visual Reasoning Beyond Language (1 citation)
- 3DMedAgent: Unified Perception-to-Understanding for 3D Medical Analysis
- 3ViewSense: Spatial and Mental Perspective Reasoning from Orthographic Views in Vision-Language Models
- Active Exploring like a Pigeon: Reinforcing Spatial Reasoning via Agentic Vision-Language Models
- ACTIVE-o3 : Empowering MLLMs with Active Perception via Pure Reinforcement Learning
- ActiveScope: Actively Seeking and Correcting Perception for MLLMs
- Adaptive Residual-Update Steering for Low-Overhead Hallucination Mitigation in Large Vision-Language Models
- Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning
- Are VLMs Seeing or Just Saying? Uncovering the Illusion of Visual Re-examination
- Artemis: Structured Visual Reasoning for Perception Policy Learning
- ASRU: Activation Steering Meets Reinforcement Unlearning for Multimodal Large Language Models
- AutoVSR: Automatic Visual-to-Symbolic Reasoning for Symbolic Expression Generation from Circuit Schematic
and 254 more in the interactive map.
Cross-modal learning · 91 papers
- OmniVideo-R1: Reinforcing Audio-visual Reasoning with Query Intention and Modality Attention (1 citation)
- UrbanFusion: Stochastic Multimodal Fusion for Contrastive Learning of Robust Spatial Representations (1 citation)
- Acoustic Interference: A New Paradigm Weaponizing Acoustic Latent Semantic for Universal Jailbreak against Large Audio Language Models
- Any2Any: Unified Arbitrary Modality Translation for Remote Sensing
- AnyMod-LLVE: Low-Light Video Enhancement with Modality-Agnostic Inference
- AOEPT: Breaking the Implicit Modality-Reduction Bottleneck in Modality-Missing Prompt Tuning
- AVTrack: Audio-Visual Tracking in Human-centric Complex Scenes
- Bimodal masked language modeling for bulk RNA-seq and DNA methylation representation learning
- Bioacoustic Geolocation: Species Sounds as Geographic Signals
- Bridging the Perceptual Gap: Residual-Enhanced Downscaling and Manifold-Aware Perception Alignment Adaptation for NR-IQA
- Calibrated Multimodal Representation Learning with Missing Modalities
- CauseCollab: Causal Unified and Modality-Agnostic Network for Heterogeneous Collaborative Perception
and 79 more in the interactive map.
Robot action and navigation · 71 papers
- AdaNav: Adaptive Reasoning with Uncertainty for Vision-Language Navigation
- AIR-VLA: Vision-Language-Action Systems for Aerial Manipulation
- AutoMoT: A Unified Vision-Language-Action Model with Asynchronous Mixture -of-Transformers for End-to-End Autonomous Driving
- Bring My Cup! Personalizing Vision-Language-Action Models with Visual Attentive Prompting
- Contrastive Representation Regularization for Vision-Language-Action Models
- DECO: Decoupled Multimodal Diffusion Transformer for Bimanual Dexterous Manipulation with a Plugin Tactile Adapter
- Decompose and Recompose: Reasoning New Skills from Existing Abilities for Cross-Task Robotic Manipulation
- Dive into the Scene: Breaking the Perceptual Bottleneck in Vision-Language Decision Making via Focus Plan Generation
- DriveWorld-VLA: Unified Latent-Space World Modeling with Vision–Language–Action for Autonomous Driving
- Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model
- DynVLA: Learning World Dynamics for Action Reasoning in Autonomous Driving
- Embodied Interpretability: Linking Causal Understanding to Generalization in Vision-Language-Action Models
and 59 more in the interactive map.
Human and embodied modeling · 70 papers
- The Geometry of Representational Failures in Vision Language Models (1 citation)
- $G^2$-Reader: Dual Evolving Graphs for Multimodal Document QA
- A Distributional View for Visual Mechanistic Interpretability: KL-Minimal Soft-Constraint Principle
- Are Object-Centric Representations Better at Compositional Generalization?
- AURA: Visually Interpretable Affective Understanding via Robust Archetypes
- Beyond Global Alignment: Fine-Grained Motion-Language Retrieval via Pyramidal Shapley-Taylor Learning
- Beyond Language Modeling: An Exploration of Multimodal Pretraining
- BrainJanus: A Unified Model for Understanding and Generation across Brain, Vision, and Language
- Can Vision Language Models Learn Intuitive Physics from Interaction?
- Closing the Loop: Universal Repository Representation with RPG-Encoder
- DeepImageSearch: Benchmarking Multimodal Agents for Context-Aware Image Retrieval in Visual Histories
- DeepSight: Long-Horizon World Modeling via Latent States Prediction for End-to-End Autonomous Driving
and 58 more in the interactive map.
CLIP and zero-shot · 40 papers
- [CLS] is Not Enough: Multi-Label Recognition via Patch-Level Inference and Adaptive Aggregation
- Advancing Analytic Class-Incremental Learning through Vision-Language Calibration
- Density-Aware Translation of Spurious Correlations in Zero-Shot VLMs
- Disease-Centric Vision-Language Pretraining with Hybrid Visual Encoding for 3D Computed Tomography
- ExpAlign: Expectation-Guided Vision–Language Alignment for Open-Vocabulary Grounding
- Geometry-Preserving Unsupervised Alignment for Heterogeneous Foundation Models
- Immuno-VLM: Immunizing Large Vision-Language Models via Generative Semantic Antibodies for Open-World Trustworthiness
- Improving CLIP Adaptation by Breaking Tail Alignment for Source-Free Cross-Domain Few-Shot Learning
- Left–Right Symmetry Breaking in CLIP-Style Vision-Language Models Trained on Synthetic Spatial-Relation Data
- LLM-Guided Diagnostic Evidence Alignment for Medical Vision–Language Pretraining under Limited Pairing
- Neutral-Reference Prompting for Vision–Language Models
- Prompt Tuning for CLIP on the Pretrained Manifold
and 28 more in the interactive map.
Visual question answering · 11 papers
- CBV: Clean-label Backdoor Attacks on Vision Language Models via Diffusion Models
- MedMosaic: A Challenging Large Scale Benchmark of Diverse Medical Audio
- MMClima: A Framework for Multimodal Climate Science Data and Evaluation
- REAL: Resolving Knowledge Conflicts in Knowledge-Intensive Visual Question Answering via Reasoning-Pivot Alignment
- ReasonEdit: Editing Vision--Language Models using Human Reasoning
- When RAG Hurts: Diagnosing and Mitigating Attention Distraction in Retrieval-Augmented LVLMs
- A Progressive Evidence Localization Framework Based on Wasserstein Gradient Flows for Document Visual Question Answering
- Benchmarking the Scientific Mind: A Pathology-Derived Biomedical VQA Benchmark for Complex Scientific Reasoning
- Beyond Single-View Indexing: Structure-Aware Multi-View Retrieval for Knowledge-Based VQA
- Dynamic Multimodal Evaluation via Knowledge-Enhanced Benchmark Evolution
- EntRAG: Entity-Centric Retrieval-Augmented Generation for Knowledge-based Visual Question Answering