STAGE-eval / README.md
boradorish's picture
Rebuild benchmark with max input tokens 3000
bab76d2 verified
|
Raw History Blame Contribute Delete
696 Bytes
metadata
dataset_info:
  features:
    - name: stem
      dtype: string
    - name: user_prompt
      dtype: string
    - name: gold_json
      dtype: string
    - name: json_schema
      dtype: string
    - name: input_tokens
      dtype: int64
    - name: user_tokens
      dtype: int64
    - name: gold_tokens
      dtype: int64
    - name: total_tokens
      dtype: int64
    - name: source_split
      dtype: string
  splits:
    - name: train
      num_bytes: 428436263
      num_examples: 18560
    - name: test
      num_bytes: 8222512
      num_examples: 851
  download_size: 112578814
  dataset_size: 436658775
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*