mlsanityData to 7 Oct 2026

Mid-Think: Training-Free Intermediate-Budget Reasoning via Token-Level Triggers

arXiv 2601.07036 ↗First posted 11 Jan 2026cs.CL cs.AI cs.LGcited by 5

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

Abstract (arXiv)

Hybrid reasoning language models are commonly controlled through high-level Think/No-think instructions to regulate reasoning behavior, yet we found that such mode switching is largely driven by a small set of trigger tokens rather than the instructions themselves. Through attention analysis and controlled prompting experiments, we show that a leading ``Okay'' token induces reasoning behavior, while the newline pattern following ``</think>'' suppresses it. Based on this observation, we propose Mid-Think, a simple training-free prompting format that combines these triggers to achieve intermediate-budget reasoning, consistently outperforming fixed-token and prompt-based baselines in terms of the accuracy-length trade-off. Furthermore, applying Mid-Think to RL training after SFT reduces training time by approximately 15% while improving final performance of Qwen3-8B on AIME from 69.8% to 72.4% and on GPQA from 58.5% to 61.1%, demonstrating its effectiveness for both inference-time control and RL-based reasoning training.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 24 of its 26 references that are in the map. First posted on arXiv 11 Jan 2026.

Builds on

24 of 26 references.

Closest ancestors · 8

Same direction, newest first

  1. 2025Demystifying Hybrid Thinking: Can LLMs Truly Switch Between Think and No-Think?14 Oct
  2. 2025Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiency10 Jun
  3. 2025Longer Context, Deeper Thinking: Uncovering the Role of Long-Context Ability in Reasoning22 May
  4. 2025Think Only When You Need with Large Hybrid-Reasoning Models20 May
  5. 2025Thinkless: LLM Learns When to Think19 May
  6. 2025Scalable Chain of Thoughts via Elastic Reasoning8 May
  7. 2025Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time12 Apr
  8. 2025LightThinker: Thinking Step-by-Step Compression21 Feb

Other references · 16

  1. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
  2. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
  3. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 1,906
  4. 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,276
  5. 2024DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Modelcited by 1,093
  6. 2024When Attention Sink Emerges in Language Models: An Empirical Viewcited by 200
  7. 2025AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning datasetcited by 165
  8. 2025Rethinking Entropy Interventions in RLVR: An Entropy Change Perspectivecited by 50
  9. 2025BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clippingcited by 47
  10. 2025Reinforcement Learning Meets Large Language Models: A Survey of Advancements and Applications Across the LLM Lifecyclecited by 34
  11. 2025Beyond GPT-5: Making LLMs Cheaper and Better via Performance-Efficiency Optimized Routingcited by 32
  12. 2025NGRPO: Negative-enhanced Group Relative Policy Optimizationcited by 31
4 more
  1. 2025Thinking Machines: A Survey of LLM based Reasoning Strategiescited by 25
  2. 2025Quagmires in SFT-RL Post-Training: When High SFT Scores Mislead and What to Use Insteadcited by 24
  3. 2025100-LongBench: Are de facto Long-Context Benchmarks Literally Evaluating Long-Context Ability?cited by 5
  4. 2025Thinking-Free Policy Initialization Makes Distilled Reasoning Models More Effective and Efficient Reasonerscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026VERGE: Formal Refinement and Guidance Engine for Verifiable LLM Reasoningcited by 11
  2. 2026Path-Lock Expert: Separating Reasoning Mode in Hybrid Thinking via Architecture-Level Separationcited by 3
  3. 2026Reliability-Gated Source Anchoring for Continual Test-Time Adaptationcited by 1
  4. 2026CausalGuard: Conformal Inference under Graph Uncertaintycited by 0
  5. 2026Agent Team Work Zone: An Automated, Persistent Workspace for Long-Lived Claude Code Agent Teamscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Fast Quiet-STaR: Thinking Without Thought Tokens
  2. 2026Batched Contextual Reinforcement: A Task-Scaling Law for Efficient Reasoning
  3. 2026HRBench: Benchmarking and Understanding Thinking-Mode Switch Strategies in Hybrid-Reasoning LLMs
  4. 2025Thinking About Thinking: Evaluating Reasoning in Post-Trained Language Models
  5. 2026Diverge to Induce Prompting: Multi-Rationale Induction for Zero-Shot Reasoning
  6. 2025Reinforcement Mid-Training
  7. 2025Effectively Controlling Reasoning Models through Thinking Intervention
  8. 2025BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens