发生了什么: The Bitter Lesson of Tool Calling

Researchers compared programmatic tool calling via Python code against JSON-based tool calling across 14 language models on the BFCL v4 benchmark.

为何重要: The Bitter Lesson of Tool Calling

Matters for engineers building LLM agents who need to choose between code-based and JSON-based tool invocation approaches for production systems.

注意: The Bitter Lesson of Tool Calling

Results are specific to BFCL v4 benchmark; real-world performance may vary depending on task complexity, tool diversity, and specific model versions deployed.

新闻

The Bitter Lesson of Tool Calling

arXiv cs.AI · 发布于 2026年8月6日 · 阅读约3分钟

30秒读懂

发生了什么: Researchers compared programmatic tool calling via Python code against JSON-based tool calling across 14 language models on the BFCL v4 benchmark.
为何重要: Matters for engineers building LLM agents who need to choose between code-based and JSON-based tool invocation approaches for production systems.
注意: Results are specific to BFCL v4 benchmark; real-world performance may vary depending on task complexity, tool diversity, and specific model versions deployed.

收听本摘要

阅读完整文章

agent
llm
language model
tool use
tool calling

The Agent Architect

每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。

每周一封邮件,一键退订。您的地址仅用于发送简报。