IDEA: Enhancing the Rule Learning Ability of Large Language Model Agent through Induction, Deduction, and Abduction

TL;DR

This paper introduces RULEARN, a benchmark for evaluating rule-learning in LLM agents within interactive environments, and proposes IDEA, a reasoning framework combining induction, deduction, and abduction to improve their rule-learning capabilities.

Contribution

The paper presents a new benchmark for rule learning in interactive settings and introduces the IDEA framework that enhances LLMs' ability to learn and apply rules in a human-like manner.

Findings

IDEA significantly improves rule-learning performance over baselines.

Discrepancies identified between human and LLM rule-learning behaviors.

RULEARN provides a challenging environment for evaluating rule learning in LLMs.

Abstract

While large language models (LLMs) have been thoroughly evaluated for deductive and inductive reasoning, their proficiency in holistic rule learning in interactive environments remains less explored. We introduce RULEARN, a novel benchmark to assess the rule-learning abilities of LLM agents in interactive settings. In RULEARN, agents strategically interact with simulated environments to gather observations, discern patterns, and solve complex problems. To enhance the rule-learning capabilities for LLM agents, we propose IDEA, a novel reasoning framework that integrates the process of Induction, Deduction, and Abduction. The IDEA agent generates initial hypotheses from limited observations through abduction, devises plans to validate these hypotheses or leverages them to solve problems via deduction, and refines previous hypotheses through induction, dynamically establishing and applying rules that mimic human rule-learning behaviors. Our evaluation of the IDEA framework, which involves five representative LLMs, demonstrates significant improvements over the baseline. Furthermore, our study with human participants reveals notable discrepancies in rule-learning behaviors between humans and LLMs. We believe our benchmark will serve as a valuable and challenging resource, and IDEA will provide crucial insights for the development of LLM agents capable of human-like rule learning in real-world scenarios. Our code and data is publicly available.