dataset_path: TIGER-Lab/MMLU-Pro
test_split: test
fewshot_split: validation
fewshot_config:
  sampler: first_n
  doc_to_text: !function utils.fewshot_to_text
  doc_to_target: ""
output_type: generate_until
doc_to_text: !function utils.doc_to_text
doc_to_target: !function utils.doc_to_target
filter_list:
  - name: "custom-extract"
    filter:
      - function: "freeform_regex"
        regex_pattern: answer is \((.*?)\)
generation_kwargs:
  until:
    - "</s>"
    - "Question:"
    - "<|im_end|>"
  max_gen_toks: 4096
  do_sample: false
  temperature: 0.0
num_fewshot: 0
metric_list:
  - metric: exact_match
    aggregation: mean
    higher_is_better: true
    ignore_case: true
    ignore_punctuation: true
metadata:
  version: 1.0
