open-athena/BFCL-RL-Campaign-Artifacts-2026.10.11

BFCL multi-harness training campaign, October 2026 This is the public, screened artifact release for the BFCL recovery campaign. The campaign did not achieve its goal of substantial BFCL-Parity improvement with cross-harness robustness. It established collection, curation, Levanter DPO, native MarinSkyRL DPO, and inline Pi evaluation infrastructure. A complete online multi-harness RL update was not demonstrated. Training and collection were stopped. BFCL-Parity is the 123-task… See the full description on the dataset page: https://huggingface.co/datasets/open-athena/BFCL-RL-Campaign-Artifacts-

<iframe src="https://silytics.si/embed/item/d439c6b0-d390-4baf-8896-efb0f5eebe5e" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.