mlsanityData to 14 Oct 2026

Aha-Flow Distillation: Flow Markers Matter in LLM Reasoning

arXiv 2609.07036 ↗First posted 7 Sep 2026cs.CL cs.AI cs.LGcited by 0

On the map: LLM reasoning → Prompting and chain-of-thought reasoning

also in 1 more
Abstract (arXiv)

We identify the Flow Moment, a reasoning pattern characterized by sustained, process-confirming verbalizations such as I'm doing, in contrast to the revision- and backtracking-oriented Aha Moment. We refer to their corresponding linguistic expressions as Flow Markers and Aha Markers, respectively. Based on this observation, we construct Flow-CoT by rewriting the discourse markers of original reasoning traces while preserving their underlying reasoning content, and use it as auxiliary supervision for on-policy self-distillation (OPSD). We further propose Aha-Flow Distillation (AFD), a dual-mode extension of OPSD that pairs different forms of privileged information with corresponding reasoning instructions. The Aha branch retains concise solution-based supervision, while the Flow branch introduces rewritten Flow-CoT under a direct and confident reasoning instruction. At inference time, the model uses only the standard reflective instruction, so Flow-style reasoning serves purely as a training signal. Experiments on AIME25 and HMMT25 show consistent improvements across Qwen3-8B and Qwen3-4B: AFD improves Avg@12 from 60.8 to 61.3 on Qwen3-8B and from 57.5 to 58.6 on Qwen3-4B over our reproduced OPSD baselines. Controlled ablations further show that, with the same Flow-CoT/Aha-CoT composition, dual-mode training improves Avg@12 from 59.5 to 60.1, indicating that the benefit comes not only from introducing heterogeneous reasoning supervision, but also from how it is organized during self-distillation. The code is available at https://github.com/Wang-Xiaodong1899/Aha-Flow-Distillation.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 22 of its 33 references that are in the map. First posted on arXiv 7 Sep 2026.

Builds on

22 of 33 references.

Closest ancestors · 7

Same direction, newest first

  1. 2026Purified OPSD: On-Policy Self-Distillation Without Losing How to Think2 Jul
  2. 2026Respecting Self-Uncertainty in On-Policy Self-Distillation for Efficient LLM Reasoning13 May
  3. 2026Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?25 Mar
  4. 2026Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models26 Jan
  5. 2025Thought Anchors: Which LLM Reasoning Steps Matter?23 Jun
  6. 2024Distilling System 2 into System 18 Jul
  7. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Models21 Mar

Other references · 15

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  4. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,210
  5. 2026DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligencecited by 1,040
  6. 2023On-Policy Distillation of Language Models: Learning from Self-Generated Mistakescited by 939
  7. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  8. 2025Demystifying Long Chain-of-Thought Reasoning in LLMscited by 326
  9. 2026MiMo-V2-Flash Technical Reportcited by 191
  10. 2026Learning beyond Teacher: Generalized On-Policy Distillation with Reward Extrapolationcited by 184
  11. 2026Entropy-Aware On-Policy Distillation of Language Modelscited by 175
  12. 2025Thoughts Are All Over the Place: On the Underthinking of o1-Like LLMscited by 163
3 more
  1. 2026Kimi K3: Open Frontier Intelligencecited by 138
  2. 2025Black-Box On-Policy Distillation of Large Language Modelscited by 59
  3. 2026PACED: Distillation and On-Policy Self-Distillation at the Frontier of Student Competencecited by 28

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026ReasoningFlow: Discourse Structures for Understanding LLM Reasoning Traces
  2. 2026Diagnosing On-Policy Self-Distillation for Reasoning Language Models
  3. 2026Backtracking When It Strays: Mitigating Dual Exposure Biases in LLM Reasoning Distillation
  4. 2026Facts Don't Speak Louder than Words: The Behavioral Essence of Long-CoT Distillation
  5. 2026Better Starts, Better Ends: Bootstrapped Iterative Self-Reasoning Distillation for Compressed Reasoning
  6. 2026FlowEdit: Information-Theoretic Control of LLM Reasoning Flows for Ill-posed Problems Involving Conflicts
  7. 2026Process Supervision via Verbal Critique Improves Reasoning in Large Language Models
  8. 2025The Challenge of Teaching Reasoning to LLMs Without RL or Distillation