eval_results.json

[
  {
    "task": "hellaswag",
    "benchmark": "HellaSwag",
    "metric": "acc_norm",
    "score": 0.26847241585341564,
    "shots": 0,
    "runtime_sec": 157.33,
    "status": "success"
  },
  {
    "task": "piqa",
    "benchmark": "PIQA",
    "metric": "acc_norm",
    "score": 0.4923830250272035,
    "shots": 0,
    "runtime_sec": 34.0,
    "status": "success"
  },
  {
    "task": "winogrande",
    "benchmark": "WinoGrande",
    "metric": "acc",
    "score": 0.4996053670086819,
    "shots": 0,
    "runtime_sec": 26.59,
    "status": "success"
  },
  {
    "task": "boolq",
    "benchmark": "BoolQ",
    "metric": "acc",
    "score": 0.6085626911314985,
    "shots": 0,
    "runtime_sec": 54.76,
    "status": "success"
  },
  {
    "task": "arc_easy",
    "benchmark": "ARC-Easy",
    "metric": "acc_norm",
    "score": 0.27104377104377103,
    "shots": 0,
    "runtime_sec": 41.15,
    "status": "success"
  },
  {
    "task": "arc_challenge",
    "benchmark": "ARC-Challenge",
    "metric": "acc_norm",
    "score": 0.27559726962457337,
    "shots": 0,
    "runtime_sec": 31.75,
    "status": "success"
  },
  {
    "task": "openbookqa",
    "benchmark": "OpenBookQA",
    "metric": "acc_norm",
    "score": 0.298,
    "shots": 0,
    "runtime_sec": 26.69,
    "status": "success"
  },
  {
    "task": "commonsense_qa",
    "benchmark": "CommonsenseQA",
    "metric": "acc",
    "score": 0.20802620802620803,
    "shots": 0,
    "runtime_sec": 30.66,
    "status": "success"
  },
  {
    "task": "lambada_openai",
    "benchmark": "LAMBADA",
    "metric": "acc",
    "score": 0.0,
    "shots": 0,
    "runtime_sec": 68.38,
    "status": "success"
  },
  {
    "task": "blimp",
    "benchmark": "BLiMP",
    "metric": "acc",
    "score": 0.536134328358209,
    "shots": 0,
    "runtime_sec": 471.77,
    "status": "success"
  },
  {
    "task": "mmlu",
    "benchmark": "MMLU",
    "metric": "acc",
    "score": 0.2416322461187865,
    "shots": 0,
    "runtime_sec": 532.07,
    "status": "success"
  },
  {
    "task": "wikitext",
    "benchmark": "WikiText-2",
    "metric": "word_perplexity",
    "score": 39301278.79233013,
    "shots": 0,
    "runtime_sec": 46.38,
    "status": "success"
  },
  {
    "task": "wikitext",
    "benchmark": "WikiText-2",
    "metric": "byte_perplexity",
    "score": 26.31431152931224,
    "shots": 0,
    "runtime_sec": 39.52,
    "status": "success"
  },
  {
    "task": "sciq",
    "benchmark": "SciQ",
    "metric": "acc_norm",
    "score": 0.2,
    "shots": 0,
    "runtime_sec": 48.87,
    "status": "success"
  },
  {
    "task": "copa",
    "benchmark": "COPA",
    "metric": "acc",
    "score": 0.63,
    "shots": 0,
    "runtime_sec": 22.29,
    "status": "success"
  },
  {
    "task": "race",
    "benchmark": "RACE",
    "metric": "acc",
    "score": 0.22392344497607655,
    "shots": 0,
    "runtime_sec": 123.26,
    "status": "success"
  },
  {
    "task": "swag",
    "benchmark": "SWAG",
    "metric": "acc_norm",
    "score": 0.2589723083075077,
    "shots": 0,
    "runtime_sec": 141.34,
    "status": "success"
  },
  {
    "task": "truthfulqa_mc2",
    "benchmark": "TruthfulQA MC2",
    "metric": "acc",
    "score": 0.4981403502526444,
    "shots": 0,
    "runtime_sec": 52.22,
    "status": "success"
  }
]