Distilling Reasoning Capabilities into Smaller Language Models

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 SO-CRATIC 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 boost 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.* Equal contribution.!LLM !Small LM Fine-tuning A robe takes 2 bolts of blue fiber and half that much white fiber.How many bolts in total does it take?" CoT: It takes 2/2=<<2/2=1>>1 bolt of white fiber.So the total amount of fabric is… Reasoning Skill TransferFew-shot Prompting Generate AnnotationCoT: How many bolts of white fiber does it take?It takes 2/2=<<2/2=1>>1 bolt of white fiber.How many bolts in total does it take?So the total amount of fabric is 2+1=<<2+1=3>>3 bolts of fabric. Socratic CoT:How many bolts of white fiber does it take?It takes… !!