How Chain-of-Thought Works? Tracing Information Flow from Decoding, Projection, and Activation
On the map: LLM reasoning → Prompting and chain-of-thought reasoning
Abstract (arXiv)
Chain-of-Thought (CoT) prompting significantly enhances model reasoning, yet its internal mechanisms remain poorly understood. We analyze CoT's operational principles by reversely tracing information flow across decoding, projection, and activation phases. Our quantitative analysis suggests that CoT may serve as a decoding space pruner, leveraging answer templates to guide output generation, with higher template adherence strongly correlating with improved performance. Furthermore, we surprisingly find that CoT modulates neuron engagement in a task-dependent manner: reducing neuron activation in open-domain tasks, yet increasing it in closed-domain scenarios. These findings offer a novel mechanistic interpretability framework and critical insights for enabling targeted CoT interventions to design more efficient and robust prompts. We released our code and data at https://anonymous.4open.science/r/cot-D247.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 14 of its 47 references that are in the map. First posted on arXiv 28 Jul 2025.
Builds on
14 of 47 references.
Closest ancestors · 6
Same direction, newest first
- 2025LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters!11 Feb
- 2024On the Hardness of Faithful Chain-of-Thought Reasoning in Large Language Models15 Jun
- 2024How to think step-by-step: A mechanistic understanding of chain-of-thought reasoning28 Feb
- 2023Contrastive Chain-of-Thought Prompting15 Nov
- 2023Analyzing Chain-of-Thought Prompting in Large Language Models via Gradient-based Feature Attributions25 Jul
- 2022Text and Patterns: For Effective Chain of Thought, It Takes Two to Tango16 Sep
Other references · 8
- 2024The Llama 3 Herd of Modelscited by 14,715
- 2021Training Verifiers to Solve Math Word Problemscited by 10,109
- 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
- 2024Concise Thoughts: Impact of Output Length on LLM Reasoning and Costcited by 109
- 2024How Do Multilingual Language Models Remember Facts?cited by 20
- 2024Towards Faithful Natural Language Explanations: A Study Using Activation Patching in Large Language Modelscited by 14
- 2024ROME: Memorization Insights from Text, Logits and Representationcited by 8
- 2025Towards Temporal Knowledge Graph Alignment in the Wildcited by 2
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2024Chain-of-Thought in Large Language Models: Decoding, Projection, and Activation
- 2025How does Chain of Thought Think? Mechanistic Interpretability of Chain-of-Thought Reasoning with Sparse Autoencoding
- 2026Mechanistic Interpretability of Chain-of-Thought Reasoning via Sequential Activation Patching
- 2026When does Chain-of-Thought Help: A Markovian Perspective
- 2024Deciphering the Factors Influencing the Efficacy of Chain-of-Thought: Probability, Memorization, and Noisy Reasoning
- 2024Understanding Reasoning in Chain-of-Thought from the Hopfieldian View
- 2025CoT is Not True Reasoning, It Is Just a Tight Constraint to Imitate: A Theory Perspective
- 2024Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worse