Efficient Long CoT Reasoning in Small Language Models
This addresses the problem of inefficient long CoT reasoning for small language models, offering an incremental improvement in distillation methods.
The paper tackles the challenge of enabling small language models (SLMs) to perform long chain-of-thought (CoT) reasoning by pruning redundant steps and using on-policy data curation, resulting in competitive performance with significantly reduced redundant steps across mathematical reasoning benchmarks.
Recent large reasoning models such as DeepSeek-R1 exhibit strong complex problems solving abilities by generating long chain-of-thought (CoT) reasoning steps. It is challenging to directly train small language models (SLMs) to emerge long CoT. Thus, distillation becomes a practical method to enable SLMs for such reasoning ability. However, the long CoT often contains a lot of redundant contents (e.g., overthinking steps) which may make SLMs hard to learn considering their relatively poor capacity and generalization. To address this issue, we propose a simple-yet-effective method to prune unnecessary steps in long CoT, and then employ an on-policy method for the SLM itself to curate valid and useful long CoT training data. In this way, SLMs can effectively learn efficient long CoT reasoning and preserve competitive performance at the same time. Experimental results across a series of mathematical reasoning benchmarks demonstrate the effectiveness of the proposed method in distilling long CoT reasoning ability into SLMs which maintains the competitive performance but significantly reduces generating redundant reasoning steps.