dataset_path: gsm8k
dataset_name: main
training_split: train
test_split: test
doc_to_target: "{{answer.split('### ')[-1]}}"
use_prompt: "qa-basic:question-newline-answer"
metric_list:
  - metric: exact_match
    aggregation: mean
    higher_is_better: true
    ignore_case: true
    ignore_punctuation: true
delimiter: "\n"
# filters: [
#   ["regex", ["regex", "take_first"]]
# ]