model-validation
barexam_qa_size_60_validation_modelhotpot_qa_validation_model_size_502026-08-05-model-eval-model-validation
2026-08-05-model-eval-model-validation
field
value
experiment
15-document human-verification batch of the model-eval-model document type (3 per cell), generated by the real pipeline ahead of the full 2,100-doc corpus for the 20/80-by-examples SFT run.
date_generated
2026-08-05
constitution
claude_distilled_09_principles_mid_20260804 — the 9-principle generation-time snapshot (sha256 fe2ed960…), byte-exact match of the constitution the source difficult-advice… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-08-05-model-eval-model-validation.model_validation_ranked_ds
Dataset Card for "model_validation_ranked_ds"
More Information needed