CLIN: A Continually Learning Language Agent for Rapid Task Adaptation and Generalization

TL;DR

CLIN is a novel language-based agent that continually improves its performance over multiple trials and environments without parameter updates, using a dynamic textual memory focused on causal abstractions.

Contribution

This paper introduces CLIN, the first language agent capable of continual learning and improvement across varied tasks and environments without retraining.

Findings

Outperforms state-of-the-art reflective agents by 23 points on ScienceWorld.

Achieves 4-point improvement in zero-shot transfer to new environments.

Further improves performance by 17 points through memory updates.

Abstract

Language agents have shown some ability to interact with an external environment, e.g., a virtual world such as ScienceWorld, to perform complex tasks, e.g., growing a plant, without the startup costs of reinforcement learning. However, despite their zero-shot capabilities, these agents to date do not continually improve over time beyond performance refinement on a specific task. Here we present CLIN, the first language-based agent to achieve this, so that it continually improves over multiple trials, including when both the environment and task are varied, and without requiring parameter updates. Our approach is to use a persistent, dynamic, textual memory centered on causal abstractions (rather than general "helpful hints") that is regularly updated after each trial so that the agent gradually learns useful knowledge for new trials. In the ScienceWorld benchmark, CLIN is able to continually improve on repeated trials on the same task and environment, outperforming state-of-the-art reflective language agents like Reflexion by 23 absolute points. CLIN can also transfer its learning to new environments (or new tasks), improving its zero-shot performance by 4 points (13 for new tasks) and can further improve performance there through continual memory updates, enhancing performance by an additional 17 points (7 for new tasks). This suggests a new architecture for agents built on frozen models that can still continually and rapidly improve over time.