mlsanityData to 7 Oct 2026

Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models

arXiv 2503.06749 ↗First posted 9 Mar 2025cs.CV cs.AI cs.CLcited by 653

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

DeepSeek-R1-Zero has successfully demonstrated the emergence of reasoning capabilities in LLMs purely through Reinforcement Learning (RL). Inspired by this breakthrough, we explore how RL can be utilized to enhance the reasoning capability of MLLMs. However, direct training with RL struggles to activate complex reasoning capabilities such as questioning and reflection in MLLMs, due to the absence of substantial high-quality multimodal reasoning data. To address this issue, we propose the reasoning MLLM, Vision-R1, to improve multimodal reasoning capability. Specifically, we first construct a high-quality multimodal CoT dataset without human annotations by leveraging an existing MLLM and DeepSeek-R1 through modality bridging and data filtering to obtain a 200K multimodal CoT dataset, Vision-R1-cold dataset. It serves as cold-start initialization data for Vision-R1. To mitigate the optimization challenges caused by overthinking after cold start, we propose Progressive Thinking Suppression Training (PTST) strategy and employ Group Relative Policy Optimization (GRPO) with the hard formatting result reward function to gradually refine the model's ability to learn correct and complex reasoning processes on a 10K multimodal math dataset. Comprehensive experiments show our model achieves an average improvement of ~6% across various multimodal math reasoning benchmarks. Vision-R1-7B achieves a 73.5% accuracy on the widely used MathVista benchmark, which is only 0.4% lower than the leading reasoning model, OpenAI O1. Scaling up the amount of multimodal math data in the RL training, Vision-R1-32B and Vison-R1-72B achieves 76.4% and 78.2% MathVista benchmark scores, respectively. The datasets and code will be released in: https://github.com/Osilly/Vision-R1 .

Key facts. Cited by 653 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 71 of its 87 references that are in the map. First posted on arXiv 9 Mar 2025.

Builds on

71 of 87 references.

Closest ancestors · 19

Same direction, newest first

  1. 2025LlamaV-o1: Rethinking Step-by-step Visual Reasoning in LLMs10 Jan
  2. 2024Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling6 Dec
  3. 2024MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scale6 Dec
  4. 2024LLaVA-CoT: Let Vision Language Models Reason Step-by-Step15 Nov
  5. 2024Improve Vision Language Model Chain-of-thought Reasoning21 Oct
  6. 2024Polymath: A Challenging Multi-modal Mathematical Reasoning Benchmark6 Oct
  7. 2024MultiMath: Bridging Visual and Mathematical Reasoning for Large Language Models30 Aug
  8. 2024Math-PUMA: Progressive Upward Multimodal Alignment to Enhance Mathematical Reasoning16 Aug
  9. 2024We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?1 Jul
  10. 2024Math-LLaVA: Bootstrapping Mathematical Reasoning for Multimodal Large Language Models25 Jun
  11. 2024TextCoT: Zoom In for Enhanced Multimodal Text-Rich Image Understanding15 Apr
  12. 2024MM-MATH: Advancing Multimodal Math Evaluation with Process Evaluation and Fine-grained Classification7 Apr
7 more
  1. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset22 Feb
  2. 2023Compositional Chain-of-Thought Prompting for Large Multimodal Models27 Nov
  3. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct
  4. 2022Super-CLEVR: A Virtual Benchmark to Diagnose Domain Robustness in Visual Reasoning1 Dec
  5. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering20 Sep
  6. 2022CLEVR-Math: A Dataset for Compositional Language, Visual and Mathematical Reasoning10 Aug
  7. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning19 Mar

Other references · 52

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  4. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  5. 2023Visual Instruction Tuningcited by 7,792
  6. 2025Qwen2.5-VL Technical Reportcited by 5,418
  7. 2024GPT-4o System Cardcited by 4,467
  8. 2023Improved Baselines with Visual Instruction Tuningcited by 4,362
  9. 2023Let's Verify Step by Stepcited by 3,598
  10. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,155
  11. 2016Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answeringcited by 2,498
  12. 2015VQA: Visual Question Answeringcited by 2,489
38 more
  1. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  2. 2019Towards VQA Models That Can Readcited by 1,696
  3. 2023MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Modelscited by 1,624
  4. 2016CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoningcited by 1,445
  5. 2024LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Modelscited by 1,442
  6. 2025s1: Simple test-time scalingcited by 1,311
  7. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,288
  8. 2023ShareGPT4V: Improving Large Multi-Modal Models with Better Captionscited by 1,000
  9. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 893
  10. 2023Graph of Thoughts: Solving Elaborate Problems with Large Language Modelscited by 873
  11. 2022Solving math word problems with process- and outcome-based feedbackcited by 871
  12. 2018VizWiz Grand Challenge: Answering Visual Questions from Blind Peoplecited by 852
  13. 2016A Diagram Is Worth A Dozen Imagescited by 793
  14. 2022A-OKVQA: A Benchmark for Visual Question Answering using World Knowledgecited by 702
  15. 2024MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?cited by 573
  16. 2023HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Modelscited by 572
  17. 2025LIMO: Less is More for Reasoningcited by 488
  18. 2023Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuningcited by 439
  19. 2021InfographicVQAcited by 430
  20. 2021Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoningcited by 376
  21. 2022Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoningcited by 347
  22. 2018DVQA: Understanding Data Visualizations via Question Answeringcited by 338
  23. 2023PMC-VQA: Visual Instruction Tuning for Medical Visual Question Answeringcited by 333
  24. 2019PlotQA: Reasoning over Scientific Plotscited by 317
  25. 2017FigureQA: An Annotated Figure Dataset for Visual Reasoningcited by 314
  26. 2023Alphazero-like Tree-Search can Guide Large Language Model Decoding and Trainingcited by 291
  27. 2024Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMscited by 265
  28. 2021IconQA: A New Benchmark for Abstract Diagram Understanding and Visual Language Reasoningcited by 262
  29. 2021GeoQA: A Geometric Question Answering Benchmark Towards Multimodal Numerical Reasoningcited by 224
  30. 2023G-LLaVA: Solving Geometric Problem with Multi-Modal Large Language Modelcited by 208
  31. 2024DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Searchcited by 189
  32. 2024Mulberry: Empowering MLLM with o1-like Reasoning and Reflection via Collective Monte Carlo Tree Searchcited by 171
  33. 2022UniGeo: Unifying Geometry Logical Reasoning via Reformulating Mathematical Expressioncited by 137
  34. 2022MultiHiertt: Numerical Reasoning over Multi Hierarchical Tabular and Textual Datacited by 120
  35. 2023GeomVerse: A Systematic Evaluation of Large Models for Geometric Reasoningcited by 107
  36. 2024Dynamic-LLaVA: Efficient Multimodal Large Language Models via Dynamic Vision-language Context Sparsificationcited by 59
  37. 2024SceMQA: A Scientific College Entrance Level Multimodal Question Answering Benchmarkcited by 35
  38. 2025MM-IQ: Benchmarking Human-Like Abstraction and Reasoning in Multimodal Modelscited by 28

50 most cited of 52 All 52 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 360; 2026: 293.

653 papers.

Sort:

By area:

  1. 2025VLM-R1: A Stable and Generalizable R1-style Large Vision-Language Modelcited by 515
  2. 2025Video-R1: Reinforcing Video Reasoning in MLLMscited by 464
  3. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Modelscited by 456
  4. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  5. 2025VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learningcited by 315
  6. 2025R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimizationcited by 296
  7. 2025Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learningcited by 267
  8. 2025GUI-R1 : A Generalist R1-Style Vision-Language Action Model For GUI Agentscited by 240
  9. 2025SFT or RL? An Early Investigation into Training R1-Like Reasoning Large Vision-Language Modelscited by 209
  10. 2025The Landscape of Agentic Reinforcement Learning for LLMs: A Surveycited by 195
  11. 2025A Survey of Reinforcement Learning for Large Reasoning Modelscited by 175
  12. 2025UI-R1: Enhancing Efficient Action Prediction of GUI Agents by Reinforcement Learningcited by 166
38 more
  1. 2025T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoTcited by 162
  2. 2025Thyme: Think Beyond Imagescited by 160
  3. 2025Cosmos-Reason1: From Physical Common Sense To Embodied Reasoningcited by 159
  4. 2025Med-R1: Reinforcement Learning for Generalizable Medical Reasoning in Vision-Language Modelscited by 144
  5. 2025SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvementcited by 144
  6. 2025OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cyclescited by 122
  7. 2025GRIT: Teaching MLLMs to Think with Imagescited by 121
  8. 2025Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Searchcited by 116
  9. 2025What Can RL Bring to VLA Generalization? An Empirical Studycited by 102
  10. 2025Latent Visual Reasoningcited by 99
  11. 2025Perception-Aware Policy Optimization for Multimodal Reasoningcited by 95
  12. 2025Self-Rewarding Vision-Language Model via Reasoning Decompositioncited by 95
  13. 2025A Sober Look at Progress in Language Model Reasoning: Pitfalls and Paths to Reproducibilitycited by 89
  14. 2025Perception, Reason, Think, and Plan: A Survey on Large Multimodal Reasoning Modelscited by 89
  15. 2025Thinking With Videos: Multimodal Tool-Augmented Reinforcement Learning for Long Video Reasoningcited by 88
  16. 2025MAI-UI Technical Report: Real-World Centric Foundation GUI Agentscited by 88
  17. 2025NoisyRollout: Reinforcing Visual Reasoning with Data Augmentationcited by 86
  18. 2025VideoRFT: Incentivizing Video Reasoning Capability in MLLMs via Reinforced Fine-Tuningcited by 86
  19. 2025Visionary-R1: Mitigating Shortcuts in Visual Reasoning with Reinforcement Learningcited by 86
  20. 2025SimpleAR: Pushing the Frontier of Autoregressive Visual Generation through Pretraining, SFT, and RLcited by 85
  21. 2025Visual Spatial Tuningcited by 85
  22. 2025VTool-R1: VLMs Learn to Think with Images via Reinforcement Learning on Multimodal Tool Usecited by 84
  23. 2025More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Modelscited by 84
  24. 2025UniVG-R1: Reasoning Guided Universal Visual Grounding with Reinforcement Learningcited by 83
  25. 2025Scaling RL to Long Videoscited by 83
  26. 2025SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Modelscited by 83
  27. 2025Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learningcited by 79
  28. 2025ARPO:End-to-End Policy Optimization for GUI Agents with Experience Replaycited by 75
  29. 2025VGR: Visual Grounded Reasoningcited by 70
  30. 2025VRAG-RL: Empower Vision-Perception-Based RAG for Visually Rich Information Understanding via Iterative Reasoning with Reinforcement Learningcited by 68
  31. 2025Ego-R1: Chain-of-Tool-Thought for Ultra-Long Egocentric Video Reasoningcited by 67
  32. 2025Traceable Evidence Enhanced Visual Grounded Reasoning: Evaluation and Methodologycited by 67
  33. 2025Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Surveycited by 66
  34. 2025R1-Reward: Training Multimodal Reward Model Through Stable Reinforcement Learningcited by 65
  35. 2025Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiencycited by 65
  36. 2025Think Only When You Need with Large Hybrid-Reasoning Modelscited by 64
  37. 2025A Survey on (M)LLM-Based GUI Agentscited by 61
  38. 2025TinyLLaVA-Video-R1: Towards Smaller LMMs for Video Reasoningcited by 59

50 most cited of 653 All 653 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Reinforced MLLM: A Survey on RL-Based Reasoning in Multimodal Large Language Models
  2. 2025LMM-R1: Empowering 3B LMMs with Strong Reasoning Abilities Through Two-Stage Rule-Based RL
  3. 2025Metis-RISE: RL Incentivizes and SFT Enhances Multimodal Reasoning Model Learning
  4. 2026DeepVision-103K: A Visually Diverse, Broad-Coverage, and Verifiable Mathematical Dataset for Multimodal Reasoning
  5. 2025R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalization
  6. 2025GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
  7. 2026Be Faithful When Response: Returning Fluent and Grounded Answers for Vision-Language Models Reinforcement Learning
  8. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning