QUALITY

NAME = QUALITY
DATASET_PATH = emozilla/quality
SAMPLE_SPLIT = validation
FEWSHOT_SPLIT = validation
RESPONSE_TYPE = LOGLIKELIHOODS
METRICS = [AccuracyLoglikelihood, AccuracyNormLoglikelihood]
SUBJECTS = ['hard', 'easy']
LANGUAGE = <Language.ENG: 'English'>

• Module: eval_framework.tasks.benchmarks.quality

• File: src/eval_framework/tasks/benchmarks/quality.py | View on GitHub

• Link to dataset: https://huggingface.co/datasets/emozilla/quality

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 "QUALITY".