{
  "cell_id": "q114_random_k2",
  "doc_id": "2023.findings-emnlp.248.pdf",
  "question": "When using GPT-4 as the base language model, how many datasets does Logic-LM (without self-refinement) outperform the two baseline models in? ",
  "answer": "4",
  "answer_format": "Int",
  "condition": "random",
  "k": 2,
  "source_position": "first",
  "modality": "image",
  "dpi": 150,
  "image_docs": [
    {
      "role": "source",
      "doc": "2023.findings-emnlp.248",
      "pages": 19,
      "image_dir": "images/2023.findings-emnlp.248"
    },
    {
      "role": "random_1",
      "doc": "web_9a644e720d72b089",
      "pages": 37,
      "image_dir": "images/web_9a644e720d72b089"
    },
    {
      "role": "random_2",
      "doc": "web_7c1cca45d1df3262",
      "pages": 19,
      "image_dir": "images/web_7c1cca45d1df3262"
    }
  ],
  "n_images": 75
}