Focused Chain-of-Thought: Efficient LLM Reasoning via Structured Input Information

TL;DR

F-CoT is a training-free, input-centric method that improves LLM reasoning efficiency by structuring essential information, reducing token use by 2-3x while maintaining accuracy on arithmetic problems.

Contribution

Introducing F-CoT, a novel structured input approach that enhances reasoning efficiency without additional training or model modifications.

Findings

Reduces token count by 2-3x on arithmetic problems

Maintains comparable accuracy to standard zero-shot CoT

Demonstrates effectiveness of structured input for efficient reasoning

Abstract

Recent large language models achieve strong reasoning performance by generating detailed chain-of-thought traces, but this often leads to excessive token use and high inference latency. Existing efficiency approaches typically focus on model-centric interventions, such as reinforcement learning or supervised fine-tuning, to reduce verbosity. In contrast, we propose a training-free, input-centric approach. Inspired by cognitive psychology, we introduce Focused Chain-of-Thought (F-CoT), which separates information extraction from the reasoning process. F-CoT first organizes the essential information from a query into a concise, structured context and then guides the model to reason exclusively over this context. By preventing attention to irrelevant details, F-CoT naturally produces shorter reasoning paths. On arithmetic word problems, F-CoT reduces generated tokens by 2-3x while maintaining accuracy comparable to standard zero-shot CoT. These results highlight structured input as a simple yet effective lever for more efficient LLM reasoning.