Tools Fail: Detecting Silent Errors in Faulty Tools
This addresses a critical reliability issue for LLM tool-use in applications like web tasks and robotics, though it appears incremental as an initial framework.
The paper tackles the problem of LLMs failing to detect silent errors when using tools, introducing a framework that explores error detection and recovery planning. Their approach shows promising results in controlled calculator and embodied agent settings.
Tools have become a mainstay of LLMs, allowing them to retrieve knowledge not in their weights, to perform tasks on the web, and even to control robots. However, most ontologies and surveys of tool-use have assumed the core challenge for LLMs is choosing the tool. Instead, we introduce a framework for tools more broadly which guides us to explore a model's ability to detect "silent" tool errors, and reflect on how to plan. This more directly aligns with the increasingly popular use of models as tools. We provide an initial approach to failure recovery with promising results both on a controlled calculator setting and embodied agent planning.