Language Model-In-The-Loop: Data Optimal Approach to Learn-To-Recommend Actions in Text Games

TL;DR

This paper investigates updating large language models during text game learning to reduce reliance on costly human annotations, showing that in-game updates can improve recommendation performance but have limited transferability across games.

Contribution

It introduces a method for updating LLMs during game learning, moving beyond fixed models and reducing dependence on annotated data.

Findings

Updating LLMs during learning improves recommendation accuracy.

Transferability of in-game trained LLMs across different games is limited.

Reducing reliance on human-annotated gameplays is feasible with in-game updates.

Abstract

Large Language Models (LLMs) have demonstrated superior performance in language understanding benchmarks. CALM, a popular approach, leverages linguistic priors of LLMs -- GPT-2 -- for action candidate recommendations to improve the performance in text games in Jericho without environment-provided actions. However, CALM adapts GPT-2 with annotated human gameplays and keeps the LLM fixed during the learning of the text based games. In this work, we explore and evaluate updating LLM used for candidate recommendation during the learning of the text based game as well to mitigate the reliance on the human annotated gameplays, which are costly to acquire. We observe that by updating the LLM during learning using carefully selected in-game transitions, we can reduce the dependency on using human annotated game plays for fine-tuning the LLMs. We conducted further analysis to study the transferability of the updated LLMs and observed that transferring in-game trained models to other games did not result in a consistent transfer.