mlsanityData to 14 Oct 2026

STOP: Structured On-Policy Pruning of Long-Form Reasoning in Low-Data Regimes

arXiv 2605.13165 ↗First posted 13 May 2026cs.CLcited by 1

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

Abstract (arXiv)

Long chain-of-thought (Long CoT) reasoning improves performance on multi-step problems, but it also induces overthinking. This inefficiency is especially problematic in low-data fine-tuning regimes, where real applications adapt reasoning models with limited supervision and cannot rely on large-scale teacher distillation or heavy test-time control. To address this, we propose STOP (Structured On-policy Pruning), an on-policy algorithm for analyzing and pruning long-form reasoning traces. STOP constructs self-distilled traces from the model. Then it maps each trace into a structured reasoning interface through node segmentation, taxonomy annotation, and reasoning-tree construction. On top of this interface, we introduce ECN (Earliest Correct Node), which retains the shortest prefix ending at the earliest node. Experiments on DeepSeek-R1-Distill-Qwen-7B and DeepSeek-R1-Distill-LLaMA-3-8B across GSM8K, Math 500, and AIME 2024 show that STOP reduces generated tokens by 19.4% to 42.4% while largely preserving accuracy in low-data fine-tuning. Beyond efficiency, our analyses show that STOP induces much smaller distributional shift than teacher-guided pruning, improves the structural efficiency of generated reasoning, and reallocates reasoning effort away from redundant verification and backtracking toward more productive exploration.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 24 of its 47 references that are in the map. First posted on arXiv 13 May 2026.

Builds on

24 of 47 references.

Closest ancestors · 10

Same direction, newest first

  1. 2025The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis25 Aug
  2. 2025Can Pruning Improve Reasoning? Revisiting Long-CoT Compression with Capability in Mind for Better Reasoning20 May
  3. 2025ShorterBetter: Guiding Reasoning Models to Find Optimal Inference Length for Efficient Reasoning30 Apr
  4. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  5. 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMs17 Feb
  6. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning13 Feb
  7. 2025Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMs30 Jan
  8. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
  9. 2024Token-Budget-Aware LLM Reasoning24 Dec
  10. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Models17 May

Other references · 14

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  2. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  3. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,210
  4. 2023On-Policy Distillation of Language Models: Learning from Self-Generated Mistakescited by 939
  5. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  6. 2025CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillationcited by 225
  7. 2025Small Models Struggle to Learn from Strong Reasonerscited by 113
  8. 2025Retaining by Doing: The Role of On-Policy Data in Mitigating Forgettingcited by 94
  9. 2024Demystifying Chains, Trees, and Graphs of Thoughtscited by 63
  10. 2025Think When You Need: Self-Adaptive Chain-of-Thought Learningcited by 38
  11. 2025NaturalThoughts: Selecting and Distilling Reasoning Traces for General Reasoning Taskscited by 32
  12. 2025What Makes a Good Reasoning Chain? Uncovering Structural Patterns in Long Chain-of-Thought Reasoningcited by 21
2 more
  1. 2025Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMscited by 19
  2. 2025Style over Substance: Distilled Language Models Reason Via Stylistic Replicationcited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AIcited by 6

Similar, not cited

Close in wording, no citation either way

  1. 2026Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models
  2. 2026To Reason but Not to Stop: Understanding and Recovering Pruned Large Reasoning Models
  3. 2026OS-Pruner: Pruning Chains-of-Thought of Reasoning Models via Optimal Stopping
  4. 2025VeriThinker: Learning to Verify Makes Reasoning Model Efficient
  5. 2025Pruning the Unsurprising: Efficient LLM Reasoning via First-Token Surprisal
  6. 2026Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning
  7. 2025Deconstructing Long Chain-of-Thought: A Structured Reasoning Optimization Framework for Long CoT Distillation
  8. 2026Graph-Based Chain-of-Thought Pruning for Reducing Redundant Reflections in Reasoning LLMs