Can Small Language Models Help Large Language Models Reason Better?: LM-Guided Chain-of-Thought
This addresses the challenge of enhancing reasoning in resource-efficient ways for AI systems, though it is incremental as it builds on existing chain-of-thought methods.
The paper tackles the problem of improving reasoning in large language models by using a small language model to generate rationales, achieving higher answer prediction accuracy on multi-hop QA benchmarks like HotpotQA and 2WikiMultiHopQA.
We introduce a novel framework, LM-Guided CoT, that leverages a lightweight (i.e., <1B) language model (LM) for guiding a black-box large (i.e., >10B) LM in reasoning tasks. Specifically, the lightweight LM first generates a rationale for each input instance. The Frozen large LM is then prompted to predict a task output based on the rationale generated by the lightweight LM. Our approach is resource-efficient in the sense that it only requires training the lightweight LM. We optimize the model through 1) knowledge distillation and 2) reinforcement learning from rationale-oriented and task-oriented reward signals. We assess our method with multi-hop extractive question answering (QA) benchmarks, HotpotQA, and 2WikiMultiHopQA. Experimental results show that our approach outperforms all baselines regarding answer prediction accuracy. We also find that reinforcement learning helps the model to produce higher-quality rationales with improved QA performance.