← Back to all news
arXiv Computation and Language By Murat Dura, Serkan \"Ozt\"urk, Selma Tekir

Mechanistic Interpretability of Chain-of-Thought Reasoning via Sequential Activation Patching

Read the original on arXiv Computation and Language →

The paper introduces a sequential activation patching framework to study how Chain-of-Thought (CoT) prompting influences large language models over multiple generated tokens. By tracking CoT-conditioned attention-head activations across token positions and aggregating them with Part-of-Speech guidance, the authors identify distributed head sets that jointly contribute to answer generation. Targeted zero-ablation experiments confirm that these heads are functionally important, affecting mechanisms such as reasoning-trajectory maintenance, answer anchoring, exemplar-target separation, and numerical generation.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computation and Language.

Related stories

arXiv Machine Learning
Jun 9

Attention Illuminates LLM Reasoning: The Preplan-and-Anchor Rhythm Enables Fine-Grained Policy Optimization

arXiv:2510. 13554v2 Announce Type: replace-cross Abstract: The reasoning pattern of Large language models (LLMs) remains opaque, and reinforcement learning (RL) typically applies uniform credit across an entire generation, blurring the distinction between pivotal and routine steps.

By Yang Li, Zhichen Dong, Yuhan Sun, Weixun Wang, Shaopan Xiong, Yijia Luo, Jiashun Liu, Han Lu, Jiamang Wang, Wenbo Su, Bo Zheng, Junchi Yan
llmsreinforcement-learning
More like this →
arXiv AI
3d ago

Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training

The paper investigates how different forms of compressed chain‑of‑thought (CoT) reasoning—Explicit, Composed, and Implicit—affect large language model (LLM) performance after supervised fine‑tuning (SFT). Using a synthetic compositional reasoning task, the authors show that coarser CoT requires more SFT data, that Composed and Implicit CoT benefit more from data scaling (with Composed also benefiting from repetition), and that reinforcement learning with verifiable rewards (RLVR) can decompose compressed steps learned during SFT. Additionally, unidirectional CoT ordering improves generalization on longer sequential tasks.

By Kohsei Matsutani, Gouki Minegishi, Takeshi Kojima, Yusuke Iwasawa, Yutaka Matsuo
llmsreinforcement-learningfine-tuning
More like this →