Hell or High Water: Evaluating Agentic Recovery from External Failures

TL;DR

This paper introduces a benchmark to evaluate how well language model agents recover from external failures during complex planning tasks, revealing current models' struggles to adapt and pursue backup strategies.

Contribution

The paper presents a novel agentic planning benchmark focused on external failures, providing systematic analysis of model performance and failure modes.

Findings

Models often identify correct functions but fail to adapt to feedback.

Performance degrades with larger search spaces and smaller models.

Scaling models improves some aspects but does not fully solve adaptation issues.

Abstract

As language model agents are applied to real world problems of increasing complexity, they will be expected to formulate plans across large search spaces. If those plans fail for reasons beyond their control, how well do language agents search for alternative ways to achieve their goals? We devise a specialized agentic planning benchmark to study this question. Each planning problem is solved via combinations of function calls. The agent searches for relevant functions from a set of over four thousand possibilities, and observes environmental feedback in the form of function outputs or error messages. Our benchmark confronts the agent with external failures in its workflow, such as functions that suddenly become unavailable. At the same time, even with the introduction of these failures, we guarantee that the task remains solvable. Ideally, an agent's performance on the planning task should not be affected by the presence of external failures. Overall, we find that language agents struggle to formulate and execute backup plans in response to environment feedback. While state-of-the-art models are often able to identify the correct function to use in the right context, they struggle to adapt to feedback from the environment and often fail to pursue alternate courses of action, even when the search space is artificially restricted. We provide a systematic analysis of the failures of both open-source and commercial models, examining the effects of search space size, as well as the benefits of scaling model size in our setting. Our analysis identifies key challenges for current generative models as well as promising directions for future work.