Self-Reasoning Language Models: Unfold Hidden Reasoning Chains with Few Reasoning Catalyst
This addresses the problem of improving reasoning in LLMs for complex tasks, offering a novel self-training approach that is incremental but shows strong gains.
The paper tackles the challenge of enhancing large language models' reasoning by introducing Self-Reasoning Language Models (SRLM), which synthesize longer chain-of-thought data through self-training, resulting in an average absolute improvement of over +2.5 points across five reasoning tasks and up to +7.89 with increased sampling.
Inference-time scaling has attracted much attention which significantly enhance the performance of Large Language Models (LLMs) in complex reasoning tasks by increasing the length of Chain-of-Thought. These longer intermediate reasoning rationales embody various meta-reasoning skills in human cognition, such as reflection and decomposition, being difficult to create and acquire. In this work, we introduce \textit{Self-Reasoning Language Model} (SRLM), where the model itself can synthesize longer CoT data and iteratively improve performance through self-training. By incorporating a few demonstration examples (i.e., 1,000 samples) on how to unfold hidden reasoning chains from existing responses, which act as a reasoning catalyst, we demonstrate that SRLM not only enhances the model's initial performance but also ensures more stable and consistent improvements in subsequent iterations. Our proposed SRLM achieves an average absolute improvement of more than $+2.5$ points across five reasoning tasks: MMLU, GSM8K, ARC-C, HellaSwag, and BBH on two backbone models. Moreover, it brings more improvements with more times of sampling during inference, such as absolute $+7.89$ average improvement with $64$ sampling times, revealing the in-depth, diverse and creative reasoning paths in SRLM against the strong baseline.