Turning Dust into Gold: Distilling Complex Reasoning Capabilities from LLMs by Leveraging Negative Data
This work addresses the challenge of making LLMs' reasoning abilities more accessible for practical applications, though it is incremental as it builds on existing distillation methods by incorporating negative data.
The paper tackles the problem of distilling reasoning capabilities from large language models (LLMs) to smaller models by leveraging negative data (incorrect reasoning chains), showing that this approach improves performance on arithmetic reasoning tasks compared to using only positive samples.
Large Language Models (LLMs) have performed well on various reasoning tasks, but their inaccessibility and numerous parameters hinder wide application in practice. One promising way is distilling the reasoning ability from LLMs to small models by the generated chain-of-thought reasoning paths. In some cases, however, LLMs may produce incorrect reasoning chains, especially when facing complex mathematical problems. Previous studies only transfer knowledge from positive samples and drop the synthesized data with wrong answers. In this work, we illustrate the merit of negative data and propose a model specialization framework to distill LLMs with negative samples besides positive ones. The framework consists of three progressive steps, covering from training to inference stages, to absorb knowledge from negative data. We conduct extensive experiments across arithmetic reasoning tasks to demonstrate the role of negative data in distillation from LLM.