WINOX_FRΒΆ
NAME = WINOX_FR
DATASET_PATH = demelin/wino_x
SAMPLE_SPLIT = test
FEWSHOT_SPLIT = test
RESPONSE_TYPE = LOGLIKELIHOODS
METRICS = [AccuracyLoglikelihood, AccuracyNormLoglikelihood]
SUBJECTS = ['lm_en_fr']
LANGUAGE = <Language.FRA: 'French'>
Module:
eval_framework.tasks.benchmarks.winoxFile: src/eval_framework/tasks/benchmarks/winox.py | View on GitHub
Link to dataset: https://huggingface.co/datasets/demelin/wino_x
More detailed documentation, with prompt examples and ground truth completions, can be generated with uv run -m eval_framework.utils.generate_task_docs --add-prompt-examples --only-tasks "WINOX_FR".