ICCV 2025 accepted papers on Vision-language and multimodal
411 of the 2,701 papers accepted at ICCV 2025 (15.2%) are on Vision-language and multimodal, compared with 7.5% at ICCV 2023. They fall into 6 topics.
See ICCV 2025 in the interactive map
Submitting to ICCV? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Multimodal LLMs · 203 papers
- Visual-RFT: Visual Reinforcement Fine-Tuning (512 citations)
- LLaVA-CoT: Let Vision Language Models Reason Step-by-Step (452 citations)
- R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalization (385 citations)
- LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models (374 citations)
- LLaVA-3D: A Simple yet Effective Pathway to Empowering LMMs with 3D Capabilities (197 citations)
- MetaMorph: Multimodal Understanding and Generation via Instruction Tuning (180 citations)
- Beyond Text-Visual Attention: Exploiting Visual Cues for Effective Token Pruning in VLMs (156 citations)
- Are VLMs Ready for Autonomous Driving? An Empirical Study from the Reliability, Data and Metric Perspectives (107 citations)
- 3DSRBench: A Comprehensive 3D Spatial Reasoning Benchmark (92 citations)
- CC-OCR: A Comprehensive and Challenging OCR Benchmark for Evaluating Large Multimodal Models in Literacy (78 citations)
- STI-Bench: Are MLLMs Ready for Precise Spatial-Temporal World Understanding? (61 citations)
- ILLUME: Illuminating Your LLMs to See, Draw, and Self-Enhance (58 citations)
and 191 more in the interactive map.
CLIP and zero-shot · 63 papers
- Derm1M: A Million-scale Vision-Language Dataset Aligned with Clinical Ontology Knowledge for Dermatology (42 citations)
- Talking to DINO: Bridging Self-Supervised Vision Backbones with Language for Open-Vocabulary Segmentation (35 citations)
- Fine-grained Abnormality Prompt Learning for Zero-shot Anomaly Detection (28 citations)
- Is CLIP ideal? No. Can we fix it? Yes! (28 citations)
- Advancing Textual Prompt Learning with Anchored Attributes (25 citations)
- CoTMR: Chain-of-Thought Multi-Scale Reasoning for Training-Free Zero-Shot Composed Image Retrieval (22 citations)
- ROADWork: A Dataset and Benchmark for Learning to Recognize, Observe, Analyze and Drive Through Work Zones (19 citations)
- Enhancing Few-Shot Vision-Language Classification with Large Multimodal Model Features (16 citations)
- WildSAT: Learning Satellite Image Representations from Wildlife Observations (16 citations)
- An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval (14 citations)
- RANKCLIP: Ranking-Consistent Language-Image Pretraining (13 citations)
- FE-CLIP: Frequency Enhanced CLIP Model for Zero-Shot Anomaly Detection and Segmentation (12 citations)
and 51 more in the interactive map.
Cross-modal learning · 63 papers
- TerraMind: Large-Scale Generative Multimodality for Earth Observation (164 citations)
- SkySense V2: A Unified Foundation Model for Multi-modal Remote Sensing (31 citations)
- XTrack: Multimodal Training Boosts RGB-X Video Object Trackers (30 citations)
- SimMLM: A Simple Framework for Multi-modal Learning with Missing Modality (27 citations)
- LUT-Fuse: Towards Extremely Fast Infrared and Visible Image Fusion via Distillation to Learnable Look-Up Tables (26 citations)
- The Scalability of Simplicity: Empirical Analysis of Vision-Language Learning with a Single Transformer (26 citations)
- Boosting Multimodal Learning via Disentangled Gradient Learning (23 citations)
- AFUNet: Cross-Iterative Alignment-Fusion Synergy for HDR Reconstruction via Deep Unfolding Paradigm (21 citations)
- Cross-modal Ship Re-Identification via Optical and SAR Imagery: A Novel Dataset and Method (19 citations)
- Improving Multimodal Learning via Imbalanced Learning (19 citations)
- Hierarchical Cross-modal Prompt Learning for Vision-Language Models (13 citations)
- Large Multi-modal Models Can Interpret Features in Large Multi-modal Models (13 citations)
and 51 more in the interactive map.
Robot action and navigation · 45 papers
- ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generation (130 citations)
- VLABench: A Large-Scale Benchmark for Language-Conditioned Robotics Manipulation with Long-Horizon Reasoning Tasks (128 citations)
- IRASim: A Fine-Grained World Model for Robot Manipulation (119 citations)
- CityNav: A Large-Scale Dataset for Real-World Aerial Navigation (75 citations)
- DrivingGPT: Unifying Driving World Modeling and Planning with Multi-modal Autoregressive Transformers (68 citations)
- Dita: Scaling Diffusion Transformer for Generalist Vision-Language-Action Policy (67 citations)
- Moto: Latent Motion Token as the Bridging Language for Learning Robot Manipulation from Videos (59 citations)
- VQ-VLA: Improving Vision-Language-Action Models via Scaling Vector-Quantized Action Tokenizers (48 citations)
- CAD-Assistant: Tool-Augmented VLLMs as Generic CAD Task Solvers (44 citations)
- NavMorph: A Self-Evolving World Model for Vision-and-Language Navigation in Continuous Environments (37 citations)
- AffordDexGrasp: Open-set Language-guided Dexterous Grasp with Generalizable-Instructive Affordance (32 citations)
- DexVLG: Dexterous Vision-Language-Grasp Model at Scale (31 citations)
and 33 more in the interactive map.
Human and embodied modeling · 29 papers
- Harmonizing Visual Representations for Unified Multimodal Understanding and Generation (77 citations)
- Towards a Unified Copernicus Foundation Model for Earth Vision (54 citations)
- SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition (46 citations)
- Where am I? Cross-View Geo-localization with Natural Language Descriptions (36 citations)
- TokenUnify: Scaling Up Autoregressive Pretraining for Neuron Segmentation (32 citations)
- MIEB: Massive Image Embedding Benchmark (19 citations)
- Beyond Walking: A Large-Scale Image-Text Benchmark for Text-based Person Anomaly Search (17 citations)
- 2.5 Years in Class: A Multimodal Textbook for Vision-Language Pretraining (14 citations)
- AerialVG: A Challenging Benchmark for Aerial Visual Grounding by Exploring Positional Relations (14 citations)
- Text-IRSTD: Leveraging Semantic Text to Promote Infrared Small Target Detection in Complex Scenes (11 citations)
- Learning Visual Hierarchies in Hyperbolic Space for Image Retrieval (8 citations)
- Modeling Human Gaze Behavior with Diffusion Models for Unified Scanpath Prediction (8 citations)
and 17 more in the interactive map.
Visual question answering · 8 papers
- GEMeX: A Large-Scale, Groundable, and Explainable Medical VQA Benchmark for Chest X-ray Diagnosis (34 citations)
- Beyond the Destination: A Novel Benchmark for Exploration-Aware Embodied Question Answering (29 citations)
- Ask and Remember: A Questions-Only Replay Strategy for Continual Visual Question Answering (8 citations)
- Acknowledging Focus Ambiguity in Visual Questions (7 citations)
- ReasonVQA: A Multi-hop Reasoning Benchmark with Structural Knowledge for Visual Question Answering (7 citations)
- AVAM: a Universal Training-free Adaptive Visual Anchoring Embedded into Multimodal Large Language Model for Multi-image Question Answering (4 citations)
- Overcoming Dual Drift for Continual Long-Tailed Visual Question Answering (1 citation)
- Towards Comprehensive Lecture Slides Understanding: Large-scale Dataset and Effective Method (1 citation)
