The Year in AI Papers
The year/Independent research

Paper 2601.06002

The Molecular Structure of Thought: Mapping the Topology of Long Chain-of-Thought Reasoning

Qiguang Chen, Yantao Du, Ziniu Li, Jinhao Liu, Songyao Duan, Jiarui Guo, Minghao Liu, Jiaheng Liu, Tong Yang, Ge Zhang, Libo Qin, Wanxiang Che, et al.

Published
Jan 2026
Research lab
Independent
Citations
3
GitHub
Not linked

01 In brief

Summary

The paper proposes that effective long chain-of-thought (Long CoT) reasoning in LLMs requires stable molecular-like structures formed by three interaction types: Deep-Reasoning (covalent-like), Self-Reflection (hydrogen-bond-like), and Self-Exploration (van der Waals-like).

Distillation from strong reasoning LLMs (e.g., DeepSeek-R1, QwQ, OpenAI-OSS) imparts these structures, while ICL from weak instruction models or human traces fails.

SFT learns reasoning structures, not surface keywords.

The authors define 'Effective Semantic Isomers' and show that only bonds promoting fast entropy convergence support stable learning; mixing incompatible structures causes structural chaos and performance drops.

They introduce Mole-Syn, a distribution-transfer-graph method that synthesizes Long CoT structures from instruction LLMs, improving performance and RL stability across six benchmarks.

Summarization and compression disrupt Long CoT structures, protecting private LLMs from distillation.

02 From the paper

Abstract

Large language models (LLMs) often fail to learn effective long chain-of-thought (Long CoT) reasoning from human or non-Long-CoT LLMs imitation. To understand this, we propose that effective and learnable Long CoT trajectories feature stable molecular-like structures in unified view, which are formed by three interaction types: Deep-Reasoning (covalent-like), Self-Reflection (hydrogen-bond-like), and Self-Exploration (van der Waals-like). Analysis of distilled trajectories reveals these structures emerge from Long CoT fine-tuning, not keyword imitation. We introduce Effective Semantic Isomers and show that only bonds promoting fast entropy convergence support stable Long CoT learning, while structural competition impairs training. Drawing on these findings, we present Mole-Syn, a distribution-transfer-graph method that guides synthesis of effective Long CoT structures, boosting performance and RL stability across benchmarks.