| group: | |
| - anli | |
| task: anli_r1 | |
| dataset_path: anli | |
| dataset_name: null | |
| output_type: multiple_choice | |
| training_split: train_r1 | |
| validation_split: dev_r1 | |
| test_split: test_r1 | |
| doc_to_text: "{{premise}}\nQuestion: {{hypothesis}} True, False, or Neither?\nAnswer:" | |
| # True = entailment | |
| # False = contradiction | |
| # Neither = neutral | |
| doc_to_target: "{{['True', 'Neither', 'False'][label]}}" | |
| doc_to_choice: | |
| - "True" | |
| - "Neither" | |
| - "False" | |
| should_decontaminate: true | |
| doc_to_decontamination_query: premise | |
| metric_list: | |
| - metric: acc | |
| aggregation: mean | |
| higher_is_better: true | |
| metadata: | |
| version: 1.0 | |