dataset_path: SaylorTwift/bbh
output_type: multiple_choice
test_split: test
doc_to_text: 'Q: {{input}}

  A:'
doc_to_target: "{{target}}"
metric_list:
  - metric: acc_norm
    aggregation: mean
    higher_is_better: true
num_fewshot: 3
fewshot_config:
  sampler: first_n
metadata:
  version: 1.0
