Darwin-31B-Opus / .eval_results /gpqa_diamond.yaml
SeaWolf-AI's picture
eval_results: describe the actual GPQA protocol (not single-sample Pass@1)
98e43e2 verified
Raw History Blame Contribute Delete
342 Bytes
- dataset:
id: Idavidrein/gpqa
task_id: diamond
value: 85.9
date: '2026-04-27'
source:
url: https://huggingface.co/FINAL-Bench/Darwin-31B-Opus
name: Model Card
notes: "Measured with the Darwin-DELPHI test-time engine on top of the model (multi-sample; protocol details proprietary). Not a single-sample Pass@1 score."