Science
DataDecide
DataDecide: OLMES multiple-choice QA evaluation of a controlled pretraining suite (25 data recipes x 14 sizes x 3 seeds). Per-instance correctness over 10 OLMES benchmarks (ARC, BoolQ, CommonsenseQA, HellaSwag, MMLU, OpenBookQA, PIQA, Social IQa, Winogrande).
37,553items
1subjects
odc-bylicense
reasoningdomain
knowledgedomain
sciencedomain
textmodality
item-level responses released
Saturation status: Unknown
Response matrix
Fit to width. Hover for subject & item; click a cell for details.
Correct (1)Incorrect (0)Unobserved
Scale: 1 = correct · 0 = incorrect
Subjects
- 1DataDecide OLMo dclm-baseline 750M seed-14 step-250000.4468