metrics/claim3_reference_metrics_partial.json

{
  "protocol": "replication of official score_bleu.py on inference JSONL P/A fields",
  "files": [
    "repro/outputs/generations/output_test_ABQ.jsonl",
    "repro/outputs/generations/output_test_AKQ.jsonl",
    "repro/outputs/generations/output_test_ALY.jsonl",
    "repro/outputs/generations/output_test_BGM.jsonl",
    "repro/outputs/generations/output_test_BOX.jsonl",
    "repro/outputs/generations/output_test_BTV.jsonl",
    "repro/outputs/generations/output_test_BUF.jsonl",
    "repro/outputs/generations/output_test_CAE.jsonl",
    "repro/outputs/generations/output_test_CAR.jsonl",
    "repro/outputs/generations/output_test_CLE.jsonl",
    "repro/outputs/generations/output_test_CTP.jsonl"
  ],
  "per_area_samples": {
    "ABQ": 46,
    "AKQ": 40,
    "ALY": 49,
    "BGM": 43,
    "BOX": 48,
    "BTV": 43,
    "BUF": 42,
    "CAE": 33,
    "CAR": 47,
    "CLE": 38,
    "CTP": 48
  },
  "total_samples": 477,
  "avg_bleu1": 0.4626,
  "avg_rouge_l": 0.3206,
  "avg_distinct1": 0.9772,
  "avg_meteor": 0.2512
}