mlsanityData to 14 Oct 2026

Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search

arXiv 2509.07969 ↗First posted 9 Sep 2025cs.CV cs.AI cs.CLcited by 120

On the map: Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning

Abstract (arXiv)

Recent advances in large multimodal models have leveraged image-based tools with reinforcement learning to tackle visual problems. However, existing open-source approaches often exhibit monotonous reasoning patterns and allow only a limited number of interaction turns, making them inadequate for difficult tasks that require trial-and-error exploration. In this work, we address this limitation by scaling up tool-based interactions and introduce Mini-o3, a system that executes deep, multi-turn reasoning -- spanning tens of steps -- and achieves state-of-the-art performance on challenging visual search tasks. Our recipe for reproducing OpenAI o3-style behaviors comprises three key components. First, we construct the Visual Probe Dataset, a collection of thousands of challenging visual search problems designed for exploratory reasoning. Second, we develop an iterative data collection pipeline to obtain cold-start trajectories that exhibit diverse reasoning patterns, including depth-first search, trial-and-error, and goal maintenance. Third, we propose an over-turn masking strategy that prevents penalization of over-turn responses (those that hit the maximum number of turns) during reinforcement learning, thereby balancing training-time efficiency with test-time scalability. Despite training with an upper bound of only six interaction turns, our model generates trajectories that naturally scale to tens of turns at inference time, with accuracy improving as the number of turns increases. Extensive experiments demonstrate that Mini-o3 produces rich reasoning patterns and deep thinking paths, effectively solving challenging visual search problems.

Key facts. Cited by 120 papers in the map (data through 14 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 38 of its 46 references that are in the map. First posted on arXiv 9 Sep 2025.

Builds on

38 of 46 references.

Closest ancestors · 1

Same direction, newest first

  1. 2022Flamingo: a Visual Language Model for Few-Shot Learning29 Apr

Other references · 37

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  2. 2017Proximal Policy Optimization Algorithmscited by 15,215
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  4. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  5. 2023Visual Instruction Tuningcited by 7,869
  6. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  7. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,137
  8. 2025Qwen2.5-VL Technical Reportcited by 5,587
  9. 2024GPT-4o System Cardcited by 4,588
  10. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,810
  11. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,719
  12. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,565
25 more
  1. 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,345
  2. 2023VILA: On Pre-training for Visual Language Modelscited by 692
  3. 2025Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Modelscited by 665
  4. 2025Visual-RFT: Visual Reinforcement Fine-Tuningcited by 513
  5. 2023V*: Guided Visual Search as a Core Mechanism in Multimodal LLMscited by 482
  6. 2025DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learningcited by 410
  7. 2025MM-Eureka: Exploring the Frontiers of Multimodal Reasoning with Rule-based Reinforcement Learningcited by 303
  8. 2025R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimizationcited by 301
  9. 2024MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?cited by 292
  10. 2025Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learningcited by 278
  11. 2025WebSailor: Navigating Super-human Reasoning for Web Agentcited by 229
  12. 2024LLaVA-UHD: an LMM Perceiving Any Aspect Ratio and High-Resolution Imagescited by 195
  13. 2025R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Modelcited by 165
  14. 2024Divide, Conquer and Combine: A Training-Free Framework for High-Resolution Image Perception in Multimodal Large Language Modelscited by 161
  15. 2025SimpleTIR: End-to-End Reinforcement Learning for Multi-Turn Tool-Integrated Reasoningcited by 146
  16. 2024Aria: An Open Multimodal Native Mixture-of-Experts Modelcited by 138
  17. 2025MMSearch-R1: Incentivizing LMMs to Searchcited by 124
  18. 2025WebShaper: Agentically Data Synthesizing via Information-Seeking Formalizationcited by 121
  19. 2025WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agentcited by 119
  20. 2025Adaptive Chain-of-Focus Reasoning via Dynamic Visual Search and Zooming for Efficient VLMscited by 96
  21. 2025Agent RL Scaling Law: Agent RL with Spontaneous Code Execution for Mathematical Problem Solvingcited by 62
  22. 2025DyFo: A Training-Free Dynamic Focus Visual Search for Enhancing LMMs in Fine-Grained Visual Understandingcited by 49
  23. 2025ACTIVE-o3: Empowering MLLMs with Active Perception via Pure Reinforcement Learningcited by 43
  24. 2025VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learningcited by 43
  25. 2025High-Resolution Visual Reasoning via Multi-Turn Grounding-Based Reinforcement Learningcited by 14

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 28; 2026: 92.

120 papers.

Sort:

By area:

  1. 2025Qwen3-VL Technical Reportcited by 2,793
  2. 2025DeepEyesV2: Toward Agentic Multimodal Modelcited by 125
  3. 2026Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillationcited by 77
  4. 2026Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perceptioncited by 50
  5. 2025TIR-Bench: A Comprehensive Benchmark for Agentic Thinking-with-Images Reasoningcited by 36
  6. 2025SenseNova-MARS: Empowering Multimodal Agentic Reasoning and Search via Reinforcement Learningcited by 33
  7. 2025JarvisEvo: Towards a Self-Evolving Photo Editing Agent with Synergistic Editor-Evaluator Optimizationcited by 31
  8. 2025Skywork-R1V4: Toward Agentic Multimodal Intelligence through Interleaved Thinking with Images and DeepResearchcited by 27
  9. 2025Skyra: AI-Generated Video Detection via Grounded Artifact Reasoningcited by 25
  10. 2026Video-o3: Native Interleaved Clue Seeking for Long Video Multi-Hop Reasoningcited by 24
  11. 2025A Survey on Agentic Multimodal Large Language Modelscited by 23
  12. 2026AgentVista: Evaluating Multimodal Agents in Ultra-Challenging Realistic Visual Scenarioscited by 23
38 more
  1. 2025AdaptVision: Efficient Vision-Language Models via Adaptive Visual Acquisitioncited by 21
  2. 2025AdaTooler-V: Adaptive Tool-Use for Images and Videoscited by 21
  3. 2025DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Modelscited by 19
  4. 2026PyVision-RL: Forging Open Agentic Vision Models via RLcited by 16
  5. 2025Thinking in 360deg: Humanoid Visual Search in the Wildcited by 15
  6. 2025Thinking with Programming Vision: Towards a Unified View for Thinking with Imagescited by 15
  7. 2026Act Wisely: Cultivating Meta-Cognitive Tool Use in Agentic Multimodal Modelscited by 14
  8. 2025ARM-Thinker: Reinforcing Multimodal Generative Reward Models with Agentic Tool Use and Visual Reasoningcited by 13
  9. 2025The Image as Its Own Reward: Reinforcement Learning with Adversarial Reward for Image Generationcited by 12
  10. 2026Imagination Helps Visual Reasoning, But Not Yet in Latent Spacecited by 12
  11. 2026VTC-Bench: Evaluating Agentic Multimodal Models via Compositional Visual Tool Chainingcited by 12
  12. 2025ViPER: Empowering the Self-Evolution of Visual Perception Abilities in Vision-Language Modelcited by 11
  13. 2025GeoVista: Web-Augmented Agentic Visual Reasoning for Geolocalizationcited by 11
  14. 2025Pathology-CoT: Learning Visual Chain-of-Thought Agent from Expert Whole Slide Image Diagnosis Behaviorcited by 10
  15. 2025Ophiuchus: Incentivizing Tool-augmented "Think with Images" for Joint Medical Segmentation, Understanding and Reasoningcited by 10
  16. 2026Thinking with Map: Reinforced Parallel Map-Augmented Agent for Geolocalizationcited by 9
  17. 2026VideoThinker: Building Agentic VideoLLMs with LLM-Guided Tool Reasoningcited by 9
  18. 2026What Does Vision Tool-Use Reinforcement Learning Really Learn? Disentangling Tool-Induced and Intrinsic Effects for Crop-and-Zoomcited by 9
  19. 2025CropVLM: Learning to Zoom for Fine-Grained Vision-Language Perceptioncited by 8
  20. 2025Revisiting the Necessity of Lengthy Chain-of-Thought in Vision-centric Reasoning Generalizationcited by 8
  21. 2025VG-Refiner: Towards Tool-Refined Referring Grounded Reasoning via Agentic Reinforcement Learningcited by 8
  22. 2025Deep But Reliable: Advancing Multi-turn Reasoning for Thinking with Imagescited by 8
  23. 2026MedSAM-Agent: Empowering Interactive Medical Image Segmentation with Multi-turn Agentic Reinforcement Learningcited by 8
  24. 2026HiVLA: A Visual-Grounded-Centric Hierarchical Embodied Manipulation Systemcited by 8
  25. 2026TerraScope: Pixel-Grounded Visual Reasoning for Earth Observationcited by 7
  26. 2026DeepScan: A Training-Free Framework for Visually Grounded Reasoning in Large Vision-Language Modelscited by 6
  27. 2026OralGPT-Plus: Learning to Use Visual Tools via Reinforcement Learning for Panoramic X-ray Analysiscited by 6
  28. 2026Q-Zoom: Query-Aware Adaptive Perception for Efficient Multimodal Large Language Modelscited by 6
  29. 2026RP-OPSD: Resolution-Privileged On-Policy Self-Distillation for Multimodal Large Language Modelscited by 6
  30. 2025Boosting Reasoning in Large Multimodal Models via Activation Replaycited by 5
  31. 2026Reliable Thinking with Imagescited by 5
  32. 2026ContextRL: Enhancing MLLM's Knowledge Discovery Efficiency with Context-Augmented RLcited by 5
  33. 2026Agentic-MME: What Agentic Capability Really Brings to Multimodal Intelligence?cited by 5
  34. 2026Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimizationcited by 5
  35. 2026Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulatorscited by 5
  36. 2026Kwai Keye-VL-2.0 Technical Reportcited by 5
  37. 2026TACO: Tool-Augmented Credit Optimization for Agentic Tool Usecited by 5
  38. 2025ERGO: Efficient High-Resolution Visual Understanding for Vision-Language Modelscited by 4

50 most cited of 120 All 120 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025InSight-o3: Empowering Multimodal Foundation Models with Generalized Visual Search
  2. 2026Visual-Seeker: Towards Visual-Native Multimodal Agentic Search via Active Visual Reasoning
  3. 2025Simple o3: Towards Interleaved Vision-Language Reasoning
  4. 2025Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning
  5. 2026Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence
  6. 2025Open-o3-Video: Grounded Video Reasoning with Explicit Spatio-Temporal Evidence
  7. 2025V-Thinker: Interactive Thinking with Images
  8. 2025V-REX: Benchmarking Exploratory Visual Reasoning via Chain-of-Questions