Distilling LLMs' Decomposition Abilities into Compact Language Models
This work addresses scalability and customization issues in AI by enabling compact models to perform complex reasoning tasks, though it is incremental as it builds on existing distillation and reinforcement learning techniques.
The study tackled the problem of compact models lacking complex reasoning abilities by distilling decomposition skills from large language models into them using offline reinforcement learning, resulting in the creation of an AI-generated dataset and baselines that demonstrate compact models can replicate such skills.
Large Language Models (LLMs) have demonstrated proficiency in their reasoning abilities, yet their large size presents scalability challenges and limits any further customization. In contrast, compact models offer customized training but often fall short in solving complex reasoning tasks. This study focuses on distilling the LLMs' decomposition skills into compact models using offline reinforcement learning. We leverage the advancements in the LLM`s capabilities to provide feedback and generate a specialized task-specific dataset for training compact models. The development of an AI-generated dataset and the establishment of baselines constitute the primary contributions of our work, underscoring the potential of compact models in replicating complex problem-solving skills.