Provably Learning from Language Feedback

TL;DR

This paper formalizes the problem of learning from language feedback, introduces a complexity measure called transfer eluder dimension, and presents a provably effective algorithm, HELiX, that outperforms naive prompting in various domains.

Contribution

It provides a theoretical framework for learning from language feedback, introduces transfer eluder dimension, and develops a no-regret algorithm with proven guarantees.

Findings

Learning from rich language feedback can be exponentially faster than from reward.

HELiX algorithm achieves performance guarantees based on transfer eluder dimension.

Empirical results show HELiX outperforms simple prompting methods.

Abstract

Interactively learning from observation and language feedback is an increasingly studied area driven by the emergence of large language model (LLM) agents. While impressive empirical demonstrations have been shown, so far a principled framing of these decision problems remains lacking. In this paper, we formalize the Learning from Language Feedback (LLF) problem, assert sufficient assumptions to enable learning despite latent rewards, and introduce $\textit{transfer eluder dimension}$ as a complexity measure to characterize the hardness of LLF problems. We show that transfer eluder dimension captures the intuition that information in the feedback changes the learning complexity of the LLF problem. We demonstrate cases where learning from rich language feedback can be exponentially faster than learning from reward. We develop a no-regret algorithm, called $\texttt{HELiX}$, that provably solves LLF problems through sequential interactions, with performance guarantees that scale with the transfer eluder dimension of the problem. Across several empirical domains, we show that $\texttt{HELiX}$ performs well even when repeatedly prompting LLMs does not work reliably. Our contributions mark a first step towards designing principled interactive learning algorithms from generic language feedback.