lm-evaluation/lm_eval/tasks/agieval/math.yaml

group:
  - agieval
  - agieval_en
task: agieval_math
dataset_path: hails/agieval-math
dataset_name: null
output_type: generate_until
training_split: null
validation_split: null
test_split: test
doc_to_text: "{{query}}"
doc_to_target: "{{answer}}"
process_results: !function utils.process_results
generation_kwargs:
  max_gen_toks: 32
  do_sample: False
  temperature: 0.0
  until:
    - "Q:"
metric_list:
  - metric: acc
    aggregation: mean
    higher_is_better: true
metadata:
  version: 1.0