Reducing Reasoning Costs: The Path of Optimization for Chain of Thought via Sparse Attention Mechanism

TL;DR

This paper introduces a sparse attention mechanism to reduce reasoning costs in large language models, demonstrating significant improvements in reasoning time and chain of thought length through experiments with GiantRabbit on linear algebra questions.

Contribution

The paper proposes a novel sparse attention mechanism and validates its effectiveness in optimizing chain of thought reasoning in large language models.

Findings

Reduced reasoning time in GiantRabbit compared to o1 Preview.

Significant decrease in chain of thought length.

Feasibility of sparse attention for reasoning optimization.

Abstract

In order to address the chain of thought in the large language model inference cost surge, this research proposes to use a sparse attention mechanism that only focuses on a few relevant tokens. The researcher constructed a new attention mechanism and used GiantRabbit trained with custom GPTs as an experimental tool. The experiment tested and compared the reasoning time, correctness score and chain of thought length of this model and o1 Preview in solving the linear algebra test questions of MIT OpenCourseWare. The results show that GiantRabbit's reasoning time and chain of thought length are significantly lower than o1 Preview. It verifies the feasibility of sparse attention mechanism for optimizing chain of thought reasoning. Detailed architectural details and experimental process have been uploaded to Github, the link is:https://github.com/brucewang123456789/GeniusTrail.git.