O3D: Offline Data-driven Discovery and Distillation for Sequential Decision-Making with Large Language Models

TL;DR

O3D leverages large-scale offline interaction data to automatically discover skills and distill knowledge, significantly enhancing LLMs' ability to solve complex sequential decision-making tasks without additional fine-tuning.

Contribution

The paper introduces O3D, a novel offline learning framework that improves LLM decision-making by discovering skills and distilling knowledge from offline data, bypassing the need for fine-tuning.

Findings

O3D improves decision-making performance on ALFWorld and WebShop benchmarks.

O3D outperforms baseline methods across various LLMs.

Offline data-driven discovery enhances LLM capabilities in complex tasks.

Abstract

Recent advancements in large language models (LLMs) have exhibited promising performance in solving sequential decision-making problems. By imitating few-shot examples provided in the prompts (i.e., in-context learning), an LLM agent can interact with an external environment and complete given tasks without additional training. However, such few-shot examples are often insufficient to generate high-quality solutions for complex and long-horizon tasks, while the limited context length cannot consume larger-scale demonstrations with long interaction horizons. To this end, we propose an offline learning framework that utilizes offline data at scale (e.g, logs of human interactions) to improve LLM-powered policies without finetuning. The proposed method O3D (Offline Data-driven Discovery and Distillation) automatically discovers reusable skills and distills generalizable knowledge across multiple tasks based on offline interaction data, advancing the capability of solving downstream tasks. Empirical results under two interactive decision-making benchmarks (ALFWorld and WebShop) verify that O3D can notably enhance the decision-making capabilities of LLMs through the offline discovery and distillation process, and consistently outperform baselines across various LLMs.