mlsanityData to 14 Oct 2026

ProxyThinker: Test-Time Guidance through Small Visual Reasoners

arXiv 2505.24872 ↗First posted 30 May 2025cs.CV cs.AI cs.CLcited by 8

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Recent advancements in reinforcement learning with verifiable rewards have pushed the boundaries of the visual reasoning capabilities in large vision-language models (LVLMs). However, training LVLMs with reinforcement fine-tuning (RFT) is computationally expensive, posing a significant challenge to scaling model size. In this work, we propose ProxyThinker, an inference-time technique that enables large models to inherit the visual reasoning capabilities from small, slow-thinking visual reasoners without any training. By subtracting the output distributions of base models from those of RFT reasoners, ProxyThinker modifies the decoding dynamics and successfully elicits the slow-thinking reasoning demonstrated by the emerged sophisticated behaviors such as self-verification and self-correction. ProxyThinker consistently boosts performance on challenging visual benchmarks on spatial, mathematical, and multi-disciplinary reasoning, enabling untuned base models to compete with the performance of their full-scale RFT counterparts. Furthermore, our implementation efficiently coordinates multiple language models with parallelism techniques and achieves up to 38 × faster inference compared to previous decoding-time methods, paving the way for the practical deployment of ProxyThinker. Code is available at https://github.com/MrZilinXiao/ProxyThinker.

Key facts. Cited by 8 papers in the map (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 55 of its 64 references that are in the map. First posted on arXiv 30 May 2025.

Builds on

55 of 64 references.

Closest ancestors · 15

Same direction, newest first

  1. 2025Bring Reason to Vision: Understanding Perception and Reasoning through Model Merging8 May
  2. 2025Skywork R1V2: Multimodal Hybrid Reinforcement Learning for Reasoning23 Apr
  3. 2025SFT or RL? An Early Investigation into Training R1-Like Reasoning Large Vision-Language Models10 Apr
  4. 2025VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning10 Apr
  5. 2025SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement10 Apr
  6. 2025VLM-R1: A Stable and Generalizable R1-style Large Vision-Language Model10 Apr
  7. 2025R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalization13 Mar
  8. 2025GTR: Guided Thought Reinforcement Prevents Thought Collapse in RL-based VLM Agent Training11 Mar
  9. 2025Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models9 Mar
  10. 2025R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Model7 Mar
  11. 2025Visual-RFT: Visual Reinforcement Fine-Tuning3 Mar
  12. 2025Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark9 Jan
3 more
  1. 2024Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling6 Dec
  2. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset22 Feb
  3. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct

Other references · 40

  1. 2017Proximal Policy Optimization Algorithmscited by 15,215
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  4. 2020Scaling Laws for Neural Language Modelscited by 6,586
  5. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  6. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  7. 2025Qwen2.5-VL Technical Reportcited by 5,587
  8. 2024GPT-4o System Cardcited by 4,588
  9. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
  10. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,810
  11. 2023MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Modelscited by 2,711
  12. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,080
28 more
  1. 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,345
  2. 2024How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suitescited by 1,050
  3. 2023mPLUG-Owl: Modularization Empowers Large Language Models with Multimodalitycited by 1,040
  4. 2025Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?cited by 982
  5. 2024Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMscited by 838
  6. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decodingcited by 688
  7. 2022Contrastive Decoding: Open-ended Text Generation as Optimizationcited by 609
  8. 2024MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?cited by 586
  9. 2023OpenFlamingo: An Open-Source Framework for Training Large Autoregressive Vision-Language Modelscited by 565
  10. 2024MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmarkcited by 512
  11. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 511
  12. 2024VLMEvalKit: An Open-Source Toolkit for Evaluating Large Multi-Modality Modelscited by 455
  13. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 441
  14. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  15. 2023DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Modelscited by 376
  16. 2025R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimizationcited by 301
  17. 2025Reinforcement Learning for Reasoning in Large Language Models with One Training Examplecited by 295
  18. 2023MIMIC-IT: Multi-Modal In-Context Instruction Tuningcited by 267
  19. 2025Crossing the Reward Bridge: Expanding RL with Verifiable Rewards Across Diverse Domainscited by 156
  20. 2024Decoding-Time Language Model Alignment with Multiple Objectivescited by 94
  21. 2024Tuning Language Models by Proxycited by 89
  22. 2023Contrastive Decoding Improves Reasoning in Large Language Modelscited by 74
  23. 2024Decoding-time Realignment of Language Modelscited by 67
  24. 2025All Roads Lead to Likelihood: The Value of Reinforcement Learning in Fine-Tuningcited by 59
  25. 2025Rethinking Reflection in Pre-Trainingcited by 43
  26. 2024Seeing the Image: Prioritizing Visual Correlation by Contrastive Alignmentcited by 20
  27. 2025Enhancing Cross-Tokenizer Knowledge Distillation with Contextual Dynamical Mappingcited by 10
  28. 2025EasySpec: Layer-Parallel Speculative Decoding for Efficient Multi-GPU Utilizationcited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 5.

8 papers.

  1. 2025OneThinker: All-in-one Reasoning Model for Image and Videocited by 56
  2. 2025AdaTooler-V: Adaptive Tool-Use for Images and Videoscited by 21
  3. 2026GlimpRouter: Efficient Collaborative Inference by Glimpsing One Token of Thoughtscited by 17
  4. 2025Reinforcement Learning for Large Model: A Surveycited by 9
  5. 2026ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decodingcited by 7
  6. 2026Low-rank Optimization Trajectories Modeling for LLM RLVR Accelerationcited by 4
  7. 2026SoftmaxGRPO: Learning to Reason using Softmax Advantage Group Estimationcited by 1
  8. 2026FRISM: Fine-Grained Reasoning Injection via Subspace-Level Model Merging for Vision-Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025More Thought, Less Accuracy? On the Dual Nature of Reasoning in Vision-Language Models
  2. 2025Look Again, Think Slowly: Enhancing Visual Reflection in Vision-Language Models
  3. 2026Addressing Overthinking in Large Vision-Language Models via Gated Perception-Reasoning Optimization
  4. 2025OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cycles
  5. 2025Reason-RFT: Reinforcement Fine-Tuning for Visual Reasoning of Vision Language Models
  6. 2026What does RL improve for Visual Reasoning? A Frankenstein-Style Analysis
  7. 2025Activating Visual Context and Commonsense Reasoning through Masked Prediction in VLMs
  8. 2025Socratic-MCTS: Test-Time Visual Reasoning by Asking the Right Questions