group: blimp
dataset_path: blimp
output_type: multiple_choice
validation_split: train
doc_to_text: ""
doc_to_target: 0
doc_to_choice: "{{[sentence_good, sentence_bad]}}"
num_fewshot: 0
should_decontaminate: true
doc_to_decontamination_query: "{{sentence_good}} {{sentence_bad}}"
metric_list:
  - metric: acc
metadata:
  version: 1.0