mlsanityData to 14 Oct 2026

Switch-Reasoner: Learn When to Think in Multitask Mixtures via Reinforcement Learning

arXiv 2607.08572 ↗First posted 9 Jul 2026cs.CVcited by 0

On the map: LLM reasoning → Test-time compute and efficient thinking

also in 2 more
Abstract (arXiv)

Multimodal Large Language Models (MLLMs) often follow a fixed Think-then-Answer paradigm, which is inefficient in heterogeneous multitask settings because simple inputs may not require explicit reasoning while difficult ones can benefit substantially from it. Learning when to think is also unstable during post-training, where imbalanced rollouts can drive the model toward always-thinking or always-direct behavior. We propose Switch-Reasoner, a GRPO-based framework that learns to adaptively select reasoning modes for MLLMs. It treats thinking as a virtual tool invocation and allows the model to either answer directly or invoke explicit reasoning before answering. To stabilize this decision, we introduce a dual-level regulation mechanism that balances the overall use of Thinking Mode and Direct Mode while providing sample-level supervision based on the relative benefit of the two choices. Experiments on 11 multimodal tasks show that Switch-Reasoner reduces unnecessary reasoning while maintaining strong performance, achieving a better accuracy-efficiency trade-off.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 54 of its 60 references that are in the map. First posted on arXiv 9 Jul 2026.

Builds on

54 of 60 references.

Closest ancestors · 10

Same direction, newest first

  1. 2026TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning13 Mar
  2. 2025ARES: Multimodal Adaptive Reasoning via Difficulty-Aware Token-Level Entropy Shaping9 Oct
  3. 2025Think in Blocks: Adaptive Reasoning from Direct Response to Deep Reasoning21 Aug
  4. 2025AdaCtrl: Towards Adaptive and Controllable Reasoning via Difficulty-Aware Budgeting24 May
  5. 2025AdaptThink: Reasoning Models Can Learn When to Think19 May
  6. 2025Thinkless: LLM Learns When to Think19 May
  7. 2025AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learning17 May
  8. 2025Learning When to Think: Shaping Adaptive Reasoning in R1-Style Models via Multi-Stage RL16 May
  9. 2025Dynamic Early Exit in Reasoning Models22 Apr
  10. 2025ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning2 Apr

Other references · 44

  1. 2017Proximal Policy Optimization Algorithmscited by 15,215
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  4. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  5. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
  6. 2025Qwen3-VL Technical Reportcited by 2,793
  7. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,719
  8. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,565
  9. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,614
  10. 2022ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoningcited by 1,479
  11. 2020DocVQA: A Dataset for VQA on Document Imagescited by 1,317
  12. 2024OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problemscited by 1,136
32 more
  1. 2024Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMscited by 838
  2. 2025Visual-RFT: Visual Reinforcement Fine-Tuningcited by 513
  3. 2025Video-R1: Reinforcing Video Reasoning in MLLMscited by 484
  4. 2021Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoningcited by 383
  5. 2023Monkey: Image Resolution and Text Label Are Important Things for Large Multi-modal Modelscited by 357
  6. 2025R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimizationcited by 301
  7. 2021IconQA: A New Benchmark for Abstract Diagram Understanding and Visual Language Reasoningcited by 267
  8. 2025VideoChat-R1: Enhancing Spatio-Temporal Perception via Reinforcement Fine-Tuningcited by 255
  9. 2025When More is Less: Understanding Chain-of-Thought Length in LLMscited by 183
  10. 2025R1-Omni: Explainable Omni-Multimodal Emotion Recognition with Reinforcement Learningcited by 100
  11. 2024Group Robust Preference Optimization in Reward-free RLHFcited by 95
  12. 2023EmoSet: A Large-scale Visual Emotion Dataset with Rich Attributescited by 82
  13. 2025A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluationscited by 67
  14. 2024PuzzleVQA: Diagnosing Multimodal Reasoning Challenges of Language Models with Abstract Visual Patternscited by 57
  15. 2025R1-ShareVL: Incentivizing Reasoning Capability of Multimodal Large Language Models via Share-GRPOcited by 52
  16. 2025Reinforced MLLM: A Survey on RL-Based Reasoning in Multimodal Large Language Modelscited by 44
  17. 2025VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learningcited by 43
  18. 2025Fast-Slow Thinking GRPO for Large Vision-Language Model Reasoningcited by 36
  19. 2025Reinforcement Learning Meets Large Language Models: A Survey of Advancements and Applications Across the LLM Lifecyclecited by 35
  20. 2025Boosting MLLM Reasoning with Text-Debiased Hint-GRPOcited by 32
  21. 2025Think or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Modelscited by 30
  22. 2025AdaTooler-V: Adaptive Tool-Use for Images and Videoscited by 21
  23. 2026VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering Twicecited by 21
  24. 2025R-4B: Incentivizing General-Purpose Auto-Thinking Capability in MLLMs via Bi-Mode Annealing and Reinforce Learningcited by 18
  25. 2025Learning to Inference Adaptively for Multimodal Large Language Modelscited by 12
  26. 2025SafeGRPO: Self-Rewarded Multimodal Safety Alignment via Rule-Governed Policy Optimizationcited by 12
  27. 2026Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Taskscited by 9
  28. 2025Mixture-of-Visual-Thoughts: Exploring Context-Adaptive Reasoning Mode Selection for General Visual Reasoningcited by 7
  29. 2025Omni-AutoThink: Adaptive Multimodal Reasoning via Reinforcement Learningcited by 5
  30. 2026EMO-R3: Reflective Reinforcement Learning for Emotional Reasoning in Multimodal Large Language Modelscited by 5
  31. 2026Learning Adaptive Reasoning Paths for Efficient Visual Reasoningcited by 5
  32. 2026AdaptR1: Reinforcement Learning Based Adaptive Interleaved Thinking in Multi-hop Question Answeringcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Advancing Multimodal Reasoning via Reinforcement Learning with Cold Start
  2. 2025ThinkDial: An Open Recipe for Controlling Reasoning Effort in Large Language Models
  3. 2025Thinking About Thinking: Evaluating Reasoning in Post-Trained Language Models
  4. 2025SAIL-RL: Guiding MLLMs in When and How to Think via Dual-Reward RL Tuning
  5. 2025Think Natively: Unlocking Multilingual Reasoning with Consistency-Enhanced Reinforcement Learning
  6. 2025GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning
  7. 2025Advancing Multimodal Reasoning: From Optimized Cold Start to Staged Reinforcement Learning
  8. 2025From Efficiency to Adaptivity: A Deeper Look at Adaptive Reasoning in Large Language Models