test:
- dataset_name: gsm8k
  dataset_type: text
  split_type: test
  metric_type: perplexity
  value: 1.726
  format: ''
  primary: true