Specializing Smaller Language Models towards Multi-Step Reasoning

TL;DR

This paper demonstrates that small language models can be specialized to perform multi-step reasoning effectively by distilling abilities from larger models, focusing their capacity on specific tasks to improve performance.

Contribution

It introduces model specialization for small models, showing how to enhance multi-step reasoning by distilling from large models and optimizing design choices.

Findings

Specialized small models outperform general small models in multi-step reasoning.

There is a tradeoff between general ability and task-specific performance.

Effective distillation can transfer complex reasoning skills to smaller models.

Abstract

The surprising ability of Large Language Models (LLMs) to perform well on complex reasoning with only few-shot chain-of-thought prompts is believed to emerge only in very large-scale models (100+ billion parameters). We show that such abilities can, in fact, be distilled down from GPT-3.5 ($\ge$ 175B) to T5 variants ($\le$ 11B). We propose model specialization, to specialize the model's ability towards a target task. The hypothesis is that large models (commonly viewed as larger than 100B) have strong modeling power, but are spread on a large spectrum of tasks. Small models (commonly viewed as smaller than 10B) have limited model capacity, but if we concentrate their capacity on a specific target task, the model can achieve a decent improved performance. We use multi-step math reasoning as our testbed because it is a very typical emergent ability. We show two important aspects of model abilities: (1). there exists a very complex balance/ tradeoff between language models' multi-dimensional abilities; (2). by paying the price of decreased generic ability, we can clearly lift up the scaling curve of models smaller than 10B towards a specialized multi-step math reasoning ability. We further give comprehensive discussions about important design choices for better generalization, including the tuning data format, the start model checkpoint, and a new model selection method. We hope our practice and discoveries can serve as an important attempt towards specialized smaller models in the new research paradigm set by LLMs.