dataset_path: JunlinHan/Multi-Level_Existence_Bench
dataset_name: "default"
test_split: train

output_type: generate_until

doc_to_visual: !function utils.mle_bench_doc_to_visual
doc_to_text: !function utils.mle_bench_doc_to_text
doc_to_target: !function utils.mle_bench_doc_to_target

process_results: !function utils.mle_bench_process_results

metric_list:
  - metric: small
    aggregation: !function utils.mle_bench_aggregate_results
    higher_is_better: true
  - metric: medium
    aggregation: !function utils.mle_bench_aggregate_results
    higher_is_better: true
  - metric: large
    aggregation: !function utils.mle_bench_aggregate_results
    higher_is_better: true
  - metric: average
    aggregation: !function utils.mle_bench_aggregate_results
    higher_is_better: true

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

lmms_eval_specific_kwargs:
  default:
    pre_prompt: ""
    post_prompt: "\nAnswer with the option's letter from the given choices directly."
  qwen_vl:
    pre_prompt: ""
    post_prompt: " Answer:"
  qwen2_5_vl:
    pre_prompt: ""
    post_prompt: "\nAnswer with the option's letter from the given choices directly."

metadata:
  - version: 0.0
