la-data / MMLU-Pro /SOURCE.md
liaialley's picture
Flatten la-data into a clean seven-dataset trajectory package
32803ff verified
|
Raw
History Blame Contribute Delete
676 Bytes
# Official evaluation setting
- Role: validation benchmark
- Source: https://github.com/TIGER-AI-Lab/MMLU-Pro
- Official API evaluator:
https://github.com/TIGER-AI-Lab/MMLU-Pro/blob/main/evaluate_from_api.py
- Official system prompt: none; the shared fallback system prompt is used.
- Applied user input: the official category-specific instruction, all official
validation demonstrations for that category, and the complete target
question and options.
- `official_few_shot_validation.parquet` is retained because those rows supply
the demonstrations; they are not pending evaluation examples.
- All 12,032 test rows are pending; earlier trajectories were discarded.