dataset_path: json
test_split: train
output_type: generate_until

generation_kwargs:
  max_new_tokens: 16
  temperature: 0
  top_p: 1.0
  num_beams: 1
  do_sample: false

metric_list:
  - metric: xmod_bench_score
    aggregation: !function utils.xmod_bench_aggregate_results
    higher_is_better: true

lmms_eval_specific_kwargs:
  default:
    post_prompt: "Answer with the option's letter (A, B, C, or D) directly."

metadata:
  version: 0.0
