KnowMap: Efficient Knowledge-Driven Task Adaptation for LLMs

TL;DR

KnowMap is a novel method that dynamically builds a knowledge base to adapt large language models to new tasks efficiently, improving performance without extensive fine-tuning.

Contribution

It introduces a knowledge-driven approach that constructs a task-specific knowledge base to enhance LLM adaptation, reducing costs and avoiding catastrophic forgetting.

Findings

Achieved 17.71% performance improvement on ScienceWorld benchmark.

Effectively integrates environmental and experiential knowledge into LLMs.

Demonstrates efficient task adaptation without extensive fine-tuning.

Abstract

While Large Language Models (LLMs) possess significant capabilities in open-world agent tasks, they also face challenges in rapidly adapting to new, specialized tasks due to their reliance on static pre-trained knowledge. Traditional methods such as fine-tuning are often costly, data-intensive, and may lead to "catastrophic forgetting." Therefore, we present KnowMap, a novel approach that dynamically constructs a knowledge base from environmental and experiential data. KnowMap fine-tunes a small knowledge-embedding model to equip a larger LLM with valuable task-specific knowledge. Our experiments on the ScienceWorld benchmark demonstrate 17.71% improvement for the performance of gpt-4-turbo model. KnowMap not only provides an efficient and effective means for LLM task-adapting, but also highlights how integrating environmental and experiential knowledge can enhance LLMs' reasoning capabilities.