training · tool-use-fine-tuning

Fine-tune on tool-call traces

Train the model on examples of deciding when to call a tool and how to use the result.

Generate or collect traces of tool calls with their arguments and results, filter for calls that helped, and fine-tune. Toolformer does this self-supervised for simple tools; ToolBench does it at scale for real APIs.

Requires: Requires training access.

Addresses: Using the tools it is given

Contexts: Coding agent, Autonomous agent

Does it work?

supported so far1 supporting · 0 contesting source · last moved 2026-09-04

Efficacy claims — what this technique actually moves, under which conditions, and whether that has been contested.

observationsingle paper
The large reported gains from fine-tuning on tool-call trajectories come from small open models starting far behind — distilling a few hundred trajectories from a stronger model moves them a long way. Whether the same procedure helps a model that is already good at tool use is a separate question this evidence does not answer.

Code

https://github.com/OpenBMB/ToolBench — Data, training recipe and evaluator for multi-tool API use.verified 2026-09-02

Does it work?

Code

Sources