ICML 2025 accepted papers on Vision-language and multimodal
178 of the 3,342 papers accepted at ICML 2025 (5.3%) are on Vision-language and multimodal, compared with 4.1% at ICML 2024. They fall into 6 topics.
See ICML 2025 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 · 57 papers
- Mitigating Object Hallucination in Large Vision-Language Models via Image-Grounded Guidance (4 citations)
- MedXpertQA: Benchmarking Expert-Level Medical Reasoning and Understanding (3 citations)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (2 citations)
- Do Vision-Language Models Really Understand Visual Language? (2 citations)
- Reasoning Limitations of Multimodal Large Language Models. A case study of Bongard Problems (2 citations)
- Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark (1 citation)
- GeoPixel: Pixel Grounding Large Multimodal Model in Remote Sensing (1 citation)
- I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models (1 citation)
- Look Twice Before You Answer: Memory-Space Visual Retracing for Hallucination Mitigation in Multimodal Large Language Models (1 citation)
- MMedPO: Aligning Medical Vision-Language Models with Clinical-Aware Multimodal Preference Optimization (1 citation)
- Parrot: Multilingual Visual Instruction Tuning (1 citation)
- SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference (1 citation)
and 45 more in the interactive map.
Human and embodied modeling · 43 papers
- HealthGPT: A Medical Large Vision-Language Model for Unifying Comprehension and Generation via Heterogeneous Knowledge Adaptation (7 citations)
- Multimodal Medical Code Tokenizer (3 citations)
- Foundation Molecular Grammar: Multi-Modal Foundation Models Induce Interpretable Molecular Graph Languages (1 citation)
- One Leaf Reveals the Season: Occlusion-Based Contrastive Learning with Semantic-Aware Views for Efficient Visual Representation (1 citation)
- OV-MER: Towards Open-Vocabulary Multimodal Emotion Recognition (1 citation)
- Beyond Cropped Regions: New Benchmark and Corresponding Baseline for Chinese Scene Text Retrieval in Diverse Layouts
- Beyond Task-Specific Reasoning: A Unified Conditional Generative Framework for Abstract Visual Reasoning
- CLIMB: Data Foundations for Large Scale Multimodal Clinical Foundation Models
- ConText: Driving In-context Learning for Text Removal and Segmentation
- Contrastive Visual Data Augmentation
- DCBM: Data-Efficient Visual Concept Bottleneck Models
- Does Data Scaling Lead to Visual Compositional Generalization?
and 31 more in the interactive map.
CLIP and zero-shot · 29 papers
- Contrastive Localized Language-Image Pre-Training (2 citations)
- EEG-Language Pretraining for Highly Label-Efficient Clinical Phenotyping (2 citations)
- QMamba: On First Exploration of Vision Mamba for Image Quality Assessment (2 citations)
- What If We Recaption Billions of Web Images with LLaMA-3? (2 citations)
- Enhancing Target-unspecific Tasks through a Features Matrix (1 citation)
- Learn from Downstream and Be Yourself in Multimodal Large Language Models Fine-Tuning (1 citation)
- An Empirical Study on Configuring In-Context Learning Demonstrations for Unleashing MLLMs' Sentimental Perception Capability
- CtrlSynth: Controllable Image Text Synthesis for Data-Efficient Multimodal Learning
- DUNIA: Pixel-Sized Embeddings via Cross-Modal Alignment for Earth Observation Applications
- Explanatory Instructions: Towards Unified Vision Tasks Understanding and Zero-shot Generalization
- From Local Details to Global Context: Advancing Vision-Language Models with Attention-Based Selection
- From Token to Rhythm: A Multi-Scale Approach for ECG-Language Pretraining
and 17 more in the interactive map.
Cross-modal learning · 26 papers
- AGAV-Rater: Adapting Large Multimodal Model for AI-Generated Audio-Visual Quality Assessment (4 citations)
- MODA: MOdular Duplex Attention for Multimodal Perception, Cognition, and Emotion Understanding (1 citation)
- A Closer Look at Multimodal Representation Collapse
- A Cross Modal Knowledge Distillation & Data Augmentation Recipe for Improving Transcriptomics Representations through Morphological Features
- AKRMap: Adaptive Kernel Regression for Trustworthy Visualization of Cross-Modal Embeddings
- Boosting Masked ECG-Text Auto-Encoders as Discriminative Learners
- Cross-Modal Alignment via Variational Copula Modelling
- FLAM: Frame-Wise Language-Audio Modeling
- FlexiReID: Adaptive Mixture of Expert for Multi-Modal Person Re-Identification
- Graph4MM: Weaving Multimodal Learning with Structural Information
- Improving Multimodal Learning Balance and Sufficiency through Data Remixing
- Learning Optimal Multimodal Information Bottleneck Representations
and 14 more in the interactive map.
Visual question answering · 13 papers
- 3D Question Answering via only 2D Vision-Language Models (1 citation)
- Watch Out Your Album! On the Inadvertent Privacy Memorization in Multi-Modal Large Language Models (1 citation)
- DocVXQA: Context-Aware Visual Explanations for Document Question Answering
- Hypo3D: Exploring Hypothetical Reasoning in 3D
- Ranked from Within: Ranking Large Multimodal Models Without Labels
- Re-ranking Reasoning Context with Tree Search Makes Large Vision-Language Models Stronger
- SK-VQA: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal LLMs
- Towards Rationale-Answer Alignment of LVLMs via Self-Rationale Calibration
- Understanding Complexity in VideoQA via Visual Program Generation
- Compositional Condition Question Answering in Tabular Understanding
- Divide and Conquer: Exploring Language-centric Tree Reasoning for Video Question-Answering
- Visual and Domain Knowledge for Professional-level Graph-of-Thought Medical Reasoning
and 1 more in the interactive map.
Robot action and navigation · 10 papers
- EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents (1 citation)
- Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models
- OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extraction
- Rethinking Latent Redundancy in Behavior Cloning: An Information Bottleneck Approach for Robot Manipulation
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation
- UI-Vision: A Desktop-centric GUI Benchmark for Visual Perception and Interaction
- UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent
- VIP: Vision Instructed Pre-training for Robotic Manipulation
- DiffusionVLA: Scaling Robot Foundation Models via Unified Diffusion and Autoregression
- Test-Time Adaptation for Online Vision-Language Navigation with Feedback-based Reinforcement Learning