Distilling Reasoning Capabilities into Smaller Language Models

TL;DR

This paper introduces Socratic CoT, a knowledge distillation method that enables small language models to acquire reasoning skills comparable to large models by decomposing problems into subproblems.

Contribution

It presents a novel distillation approach using Socratic CoT to transfer reasoning capabilities from large models to smaller ones, improving their performance significantly.

Findings

Small models outperform baselines by over 70% on reasoning tasks.

Socratic CoT enables smaller models to sometimes surpass larger models in reasoning.

Distilled models perform well across multiple reasoning datasets.

Abstract

Step-by-step reasoning approaches like chain of thought (CoT) have proved to be very effective in inducing reasoning capabilities in large language models. However, the success of the CoT approach is fundamentally tied to the model size, and billion parameter-scale models are often needed to get CoT to work. In this paper, we propose a knowledge distillation approach that leverages the step-by-step CoT reasoning capabilities of larger models and distills these abilities into smaller models. In this work, we propose an alternative reasoning scheme, Socratic CoT, that learns a decomposition of the original problem into a sequence of subproblems and uses it to guide the intermediate reasoning steps. We use Socratic CoT to train a combination of two small distilled models: a problem decomposer and a subproblem solver. In practice, given a new problem, the two distilled models work in sync to decompose and solve complex problems. On multiple reasoning datasets (GSM8K, StrategyQA, and SVAMP), our proposed distillation strategies boosts the performance of smaller models over 70% compared to the baselines. Finally, we investigate when Socratic CoT is an effective alternative to CoT, demonstrating cases where a much smaller model (GPT-2 large) can outperform a 10X larger model (GPT-3 6B). Our code is available here: https://github.com/kumar-shridhar/Distiiling-LM