Mixture of Reasonings: Teach Large Language Models to Reason with Adaptive Strategies

TL;DR

The paper introduces Mixture of Reasoning (MoR), a training framework that enables large language models to autonomously adapt their reasoning strategies across tasks, improving performance without manual prompt engineering.

Contribution

MoR is a novel training approach that embeds diverse reasoning strategies into LLMs, allowing autonomous, task-adaptive reasoning and outperforming traditional prompt-based methods.

Findings

MoR significantly improves reasoning accuracy on benchmark tasks.

MoR eliminates the need for task-specific prompt engineering.

MoR achieves up to 13.5% performance improvement over baselines.

Abstract

Large language models (LLMs) excel in complex tasks through advanced prompting techniques like Chain-of-Thought (CoT) and Tree-of-Thought (ToT), but their reliance on manually crafted, task-specific prompts limits adaptability and efficiency. We introduce Mixture of Reasoning (MoR), a training framework that embeds diverse reasoning strategies into LLMs for autonomous, task-adaptive reasoning without external prompt engineering. MoR has two phases: Thought Generation, creating reasoning chain templates with models like GPT-4o, and SFT Dataset Construction, pairing templates with benchmark datasets for supervised fine-tuning. Our experiments show that MoR significantly enhances performance, with MoR150 achieving 0.730 (2.2% improvement) using CoT prompting and 0.734 (13.5% improvement) compared to baselines. MoR eliminates the need for task-specific prompts, offering a generalizable solution for robust reasoning across diverse tasks.