Chain of Methodologies: Scaling Test Time Computation without Training

TL;DR

This paper presents the Chain of Methodologies (CoM), a prompting framework that enhances LLM reasoning by integrating human insights, enabling complex task handling without additional training, and surpassing existing baselines.

Contribution

The introduction of CoM, a training-free prompting method that incorporates human methodologies to improve structured reasoning in LLMs.

Findings

CoM outperforms competitive baselines on complex reasoning tasks.

CoM activates systematic reasoning without explicit fine-tuning.

Demonstrates potential of training-free methods for advanced reasoning.

Abstract

Large Language Models (LLMs) often struggle with complex reasoning tasks due to insufficient in-depth insights in their training data, which are typically absent in publicly available documents. This paper introduces the Chain of Methodologies (CoM), an innovative and intuitive prompting framework that enhances structured thinking by integrating human methodological insights, enabling LLMs to tackle complex tasks with extended reasoning. CoM leverages the metacognitive abilities of advanced LLMs, activating systematic reasoning throught user-defined methodologies without explicit fine-tuning. Experiments show that CoM surpasses competitive baselines, demonstrating the potential of training-free prompting methods as robust solutions for complex reasoning tasks and bridging the gap toward human-level reasoning through human-like methodological insights.