Apps · PROJECT NOTES
ToolACE Data
The public synthetic tool-calling dialogue subset released with ToolACE, containing tool schemas, user requests, assistant calls and simulated tool responses.
What can you do with it?
Train or inspect function-calling dialogue models, including multiple calls, missing parameters and irrelevant-tool responses, using system/conversations records.
What should you know before using it?
This is a public 11,300-row train subset, not the full generation pipeline, a held-out benchmark or an executable API environment. The paper's 26,507-API synthesis pool is not the released dialogue count or a verified inventory of callable services. Tool responses are synthetic training content. Associated ToolACE model releases are separate artifacts; no new evaluation score is supplied by this data entry.
What is verified here?
Runtime: Public ungated JSON dataset; Hub metadata exposes the default/train split with 11,300 rows and system/conversations columns. No API credentials or runtime are needed to inspect the data; training infrastructure must be supplied separately. The 37 MB corpus was not downloaded.
License note: The official dataset card declares Apache-2.0. The inspected repository contains README.md and data.json but no separate LICENSE file; model licenses are not inferred from the data card. Upstream license ↗
This entry was reviewed against original project material. We have not independently run the software or reproduced its results. AI-development claims are author-reported; check upstream for current versions, setup requirements, and licenses.
- Resource type
- Dataset
- Provenance
- Development provenance not established
- Reviewed
- Platforms
- Consult upstream documentation