WMT16

NAME = WMT16
DATASET_PATH = wmt16
SAMPLE_SPLIT = test
FEWSHOT_SPLIT = test
RESPONSE_TYPE = COMPLETION
METRICS = [LINEWISE_BLEU, LINEWISE_CHRF, LINEWISE_TER]
SUBJECTS = ['de-en', 'en-de']
LANGUAGE = {'de-en': (<Language.DEU: 'German'>, <Language.ENG: 'English'>), 'en-de': (<Language.ENG: 'English'>, <Language.DEU: 'German'>)}

• Module: eval_framework.tasks.benchmarks.wmt

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

• Link to dataset: https://huggingface.co/datasets/wmt16

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 "WMT16".