何が起きたか: 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.

なぜ重要か: 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.

注意点: 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.

ニュース

The Bitter Lesson of Tool Calling

arXiv cs.AI · 公開日 2026年8月6日 · 読了3分

30秒で要点

何が起きたか: Researchers compared programmatic tool calling via Python code against JSON-based tool calling across 14 language models on the BFCL v4 benchmark.
なぜ重要か: Matters for engineers building LLM agents who need to choose between code-based and JSON-based tool invocation approaches for production systems.
注意点: Results are specific to BFCL v4 benchmark; real-world performance may vary depending on task complexity, tool diversity, and specific model versions deployed.

この要約を音声で聴く

記事の全文を読む

agent
llm
language model
tool use
tool calling

The Agent Architect

1つのパターン、1つのトレードオフ、1つの本番障害事例。エージェントシステムを構築する人のための短い週刊ブリーフィング。

週1回のメール、ワンクリックで購読解除できます。アドレスはブリーフィングの送信のみに使用します。