Tag
This paper introduces a closed-world resolution method to combat tool hallucination in LLM agents, offering a taxonomy and benchmark for measuring and addressing fabricated tool calls.