classic_pythia-2.8b-step2000/groups/json/knowledge_accuracy.json

{
  "title": "Accuracy",
  "header": [
    {
      "value": "Model",
      "markdown": false,
      "metadata": {}
    },
    {
      "value": "Mean win rate",
      "description": "How many models this model outperforms on average (over columns).",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {}
    },
    {
      "value": "NaturalQuestions (closed-book) - F1",
      "description": "The NaturalQuestions [(Kwiatkowski et al., 2019)](https://aclanthology.org/Q19-1026/) benchmark for question answering based on naturally-occurring queries through Google Search. The input does not include the Wikipedia page with the answer.\n\nF1: Average F1 score in terms of word overlap between the model output and correct reference.",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {
        "metric": "F1",
        "run_group": "NaturalQuestions (closed-book)"
      }
    },
    {
      "value": "HellaSwag - EM",
      "description": "The HellaSwag benchmark for commonsense reasoning in question answering [(Zellers et al., 2019)](https://aclanthology.org/P19-1472/).\n\nExact match: Fraction of instances that the predicted output matches a correct reference exactly.",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {
        "metric": "EM",
        "run_group": "HellaSwag"
      }
    },
    {
      "value": "OpenbookQA - EM",
      "description": "The OpenbookQA benchmark for commonsense-intensive open book question answering [(Mihaylov et al., 2018)](https://aclanthology.org/D18-1260/).\n\nExact match: Fraction of instances that the predicted output matches a correct reference exactly.",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {
        "metric": "EM",
        "run_group": "OpenbookQA"
      }
    },
    {
      "value": "TruthfulQA - EM",
      "description": "The TruthfulQA benchmarking for measuring model truthfulness and commonsense knowledge in question answering [(Lin et al., 2022)](https://aclanthology.org/2022.acl-long.229/).\n\nExact match: Fraction of instances that the predicted output matches a correct reference exactly.",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {
        "metric": "EM",
        "run_group": "TruthfulQA"
      }
    },
    {
      "value": "MMLU - EM",
      "description": "The Massive Multitask Language Understanding (MMLU) benchmark for knowledge-intensive question answering across 57 domains [(Hendrycks et al., 2021)](https://openreview.net/forum?id=d7KBjmI3GmQ).\n\nExact match: Fraction of instances that the predicted output matches a correct reference exactly.",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {
        "metric": "EM",
        "run_group": "MMLU"
      }
    },
    {
      "value": "WikiFact - EM",
      "description": "Scenario introduced in this work, inspired by [Petroni et al. (2019)](https://aclanthology.org/D19-1250/), to more extensively test factual knowledge.\n\nQuasi-exact match: Fraction of instances that the predicted output matches a correct reference up to light processing.",
      "markdown": false,
      "lower_is_better": false,
      "metadata": {
        "metric": "EM",
        "run_group": "WikiFact"
      }
    }
  ],
  "rows": [
    [
      {
        "value": "EleutherAI/pythia-2.8b",
        "description": "",
        "markdown": false
      },
      {
        "markdown": false
      },
      {
        "description": "No matching runs",
        "markdown": false
      },
      {
        "description": "No matching runs",
        "markdown": false
      },
      {
        "description": "No matching runs",
        "markdown": false
      },
      {
        "description": "No matching runs",
        "markdown": false
      },
      {
        "value": 0.27487719298245616,
        "description": "min=0.16, mean=0.275, max=0.42, sum=1.374 (5)",
        "style": {
          "font-weight": "bold"
        },
        "markdown": false,
        "run_spec_names": [
          "mmlu:subject=abstract_algebra,method=multiple_choice_joint,model=EleutherAI_pythia-2.8b,data_augmentation=canonical",
          "mmlu:subject=college_chemistry,method=multiple_choice_joint,model=EleutherAI_pythia-2.8b,data_augmentation=canonical",
          "mmlu:subject=computer_security,method=multiple_choice_joint,model=EleutherAI_pythia-2.8b,data_augmentation=canonical",
          "mmlu:subject=econometrics,method=multiple_choice_joint,model=EleutherAI_pythia-2.8b,data_augmentation=canonical",
          "mmlu:subject=us_foreign_policy,method=multiple_choice_joint,model=EleutherAI_pythia-2.8b,data_augmentation=canonical"
        ]
      },
      {
        "value": 0.0075921845770110154,
        "description": "min=0, mean=0.008, max=0.034, sum=0.076 (10)",
        "style": {
          "font-weight": "bold"
        },
        "markdown": false,
        "run_spec_names": [
          "wikifact:k=5,subject=author,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=currency,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=discoverer_or_inventor,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=instance_of,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=medical_condition_treated,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=part_of,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=place_of_birth,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=plaintiff,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=position_held,model=EleutherAI_pythia-2.8b",
          "wikifact:k=5,subject=symptoms_and_signs,model=EleutherAI_pythia-2.8b"
        ]
      }
    ]
  ],
  "links": [
    {
      "text": "LaTeX",
      "href": "benchmark_output/runs/classic_pythia-2.8b-step2000/groups/latex/knowledge_accuracy.tex"
    },
    {
      "text": "JSON",
      "href": "benchmark_output/runs/classic_pythia-2.8b-step2000/groups/json/knowledge_accuracy.json"
    }
  ],
  "name": "accuracy"
}