aiqbench analysis
This study demystifies Chain-of-Thought prompting by revealing how it functions as a decoding space pruner that leverages answer templates to guide generation. The research further uncovers task-dependent modulation of neuron engagement, showing reduced activation in open-domain tasks versus increased activity in closed-domain scenarios. These mechanistic insights provide a crucial framework for designing more efficient and robust AI prompts through targeted interventions.
How Chain-of-Thought Works? Tracing Information Flow from Decoding, Projection, and Activation
arXiv:2507.20758v3 Announce Type: replace Abstract: 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.
Title:How Chain-of-Thought Works? Tracing Information Flow from Decoding, Projection, and Activation
Abstract: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 this https URL.
Artificial Intelligence (cs.AI)
arXiv:2507.20758 [cs.AI]
(or arXiv:2507.20758v3 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2507.20758 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 220 words
