lm-evaluation-harness/lm_eval/tasks/super_glue/multirc/default.yaml

group:
  - super-glue-lm-eval-v1
task: multirc
dataset_path: super_glue
dataset_name: multirc
output_type: multiple_choice
training_split: train
validation_split: validation
doc_to_text: "{{paragraph}}\nQuestion: {{question}}\nAnswer:"
doc_to_target: label
doc_to_choice: "['''{{answer}}\\nIs the answer correct? yes''', '''{{answer}}\\nIs the answer correct? no''']"
metric_list:
  - metric: acc
metadata:
  version: 2.0