ToolTalk: Evaluating Tool-Usage in a Conversational Setting
This addresses the problem of measuring tool-augmented AI assistants for researchers and developers, though it is incremental as it builds on existing benchmarking efforts.
The paper introduces ToolTalk, a benchmark for evaluating large language model assistants in conversational settings that require multi-step tool usage, achieving success rates of 26% for GPT-3.5 and 50% for GPT-4.
Large language models (LLMs) have displayed massive improvements in reasoning and decision-making skills and can hold natural conversations with users. Many recent works seek to augment LLM-based assistants with external tools so they can access private or up-to-date information and carry out actions on behalf of users. To better measure the performance of these assistants, this paper introduces ToolTalk, a benchmark consisting of complex user intents requiring multi-step tool usage specified through dialogue. ToolTalk contains 28 tools grouped into 7 plugins, and includes a complete simulated implementation of each tool, allowing for fully automated evaluation of assistants that rely on execution feedback. ToolTalk also emphasizes tools that externally affect the world rather than only tools for referencing or searching information. We evaluate GPT-3.5 and GPT-4 on ToolTalk resulting in success rates of 26% and 50% respectively. Our analysis of the errors reveals three major categories and suggests some future directions for improvement. We release ToolTalk at https://github.com/microsoft/ToolTalk.