Meta 训练 8B 模型做工具选择，万级工具下超过 402B 大模型
Meta 把 Qwen3 8B 训成专用工具选择器，一万多个工具的场景下比 402B 大模型选得还准，做法和数字都在这篇里。
Meta 的研究针对超过一万个工具场景下的 Agent 工具选择问题，此时工具描述无法全部放进上下文。他们基于 Plan, Retrieve, Select 流程训练 Qwen3 模型：先把指令拆成子步骤，检索器为每步返回候选，模型再逐步选定工具。训练先用约 2 万条 ToolBench 指令做 SFT，再用按黄金工具覆盖数和排名打分的奖励做 RL。在 ToolBench G3 上，8B 模型从 SFT 后的 36.2% 提升到 RL 后的 41.5%，高于被提示的 Llama-4-Maverick，也优于 402B 模型。
Great application of small language models by Meta.
They find that an 8B model picks tools better than a 402B model when trained for the job.
This work focuses on tool selection for agents with more than ten thousand tools.
At that scale, tool descriptions won't fit in context, and a retriever can only rank candidates without knowing which tools the instruction actually needs. Many systems hand this step to a proprietary LLM.
They train Qwen3 models on a Plan, Retrieve, Select workflow. The model splits the instruction into substeps, the retriever returns candidates for each one, and the model commits one tool per substep. Training is SFT on about 20K curated ToolBench instructions, then RL with a reward that scores how many gold tools are covered and at what rank.
On ToolBench G3, the 8B model goes from 36.2% after SFT to 41.5% after RL, above prompted Llama-4-Maverick.
The trained selector also works with different retrievers and on tool corpora from new domains.
https://t.co/PoKsL4iqkG
Chat with Paper: https://t.co/6GMrNnk4tL
- Artificial Analysis10-21 17:35原文
