[
  {
    "id": 0,
    "task": "dense_captioning_gemini",
    "data_source": "AVOS",
    "question": "You are an expert in interpreting surgical videos. This video clip depicts an open surgery. Construct a compact summary of all occurrences of ['cutting', 'tying', 'suturing']. For each, provide [start, end] and a one-sentence caption.",
    "frames": [
      "live_examples/live_00/frame_000.jpg",
      "live_examples/live_00/frame_001.jpg",
      "live_examples/live_00/frame_002.jpg",
      "live_examples/live_00/frame_003.jpg",
      "live_examples/live_00/frame_004.jpg",
      "live_examples/live_00/frame_005.jpg",
      "live_examples/live_00/frame_006.jpg",
      "live_examples/live_00/frame_007.jpg",
      "live_examples/live_00/frame_008.jpg",
      "live_examples/live_00/frame_009.jpg",
      "live_examples/live_00/frame_010.jpg",
      "live_examples/live_00/frame_011.jpg",
      "live_examples/live_00/frame_012.jpg",
      "live_examples/live_00/frame_013.jpg",
      "live_examples/live_00/frame_014.jpg",
      "live_examples/live_00/frame_015.jpg",
      "live_examples/live_00/frame_016.jpg",
      "live_examples/live_00/frame_017.jpg",
      "live_examples/live_00/frame_018.jpg",
      "live_examples/live_00/frame_019.jpg",
      "live_examples/live_00/frame_020.jpg",
      "live_examples/live_00/frame_021.jpg",
      "live_examples/live_00/frame_022.jpg",
      "live_examples/live_00/frame_023.jpg",
      "live_examples/live_00/frame_024.jpg",
      "live_examples/live_00/frame_025.jpg",
      "live_examples/live_00/frame_026.jpg",
      "live_examples/live_00/frame_027.jpg",
      "live_examples/live_00/frame_028.jpg",
      "live_examples/live_00/frame_029.jpg",
      "live_examples/live_00/frame_030.jpg",
      "live_examples/live_00/frame_031.jpg",
      "live_examples/live_00/frame_032.jpg",
      "live_examples/live_00/frame_033.jpg",
      "live_examples/live_00/frame_034.jpg",
      "live_examples/live_00/frame_035.jpg",
      "live_examples/live_00/frame_036.jpg",
      "live_examples/live_00/frame_037.jpg",
      "live_examples/live_00/frame_038.jpg",
      "live_examples/live_00/frame_039.jpg",
      "live_examples/live_00/frame_040.jpg",
      "live_examples/live_00/frame_041.jpg",
      "live_examples/live_00/frame_042.jpg",
      "live_examples/live_00/frame_043.jpg",
      "live_examples/live_00/frame_044.jpg",
      "live_examples/live_00/frame_045.jpg",
      "live_examples/live_00/frame_046.jpg",
      "live_examples/live_00/frame_047.jpg",
      "live_examples/live_00/frame_048.jpg",
      "live_examples/live_00/frame_049.jpg",
      "live_examples/live_00/frame_050.jpg",
      "live_examples/live_00/frame_051.jpg",
      "live_examples/live_00/frame_052.jpg"
    ],
    "n_frames": 53,
    "test_idx": 261
  },
  {
    "id": 1,
    "task": "dense_captioning_gpt",
    "data_source": "CholecT50",
    "question": "You are an expert in operative video analysis. The following video contains endoscopic footage of a laparoscopic cholecystectomy, where the gallbladder is removed using small incisions and specialized instruments. Detect and caption only clinically relevant actions from ['preparation', 'carlot-triangle-dissection', 'clipping-and-cutting', 'gallbladder-dissection', 'gallbladder-packaging', 'cleaning-and-coagulation', 'gallbladder-extraction']. Return start/end times with a brief, objective description.",
    "frames": [
      "live_examples/live_01/frame_000.jpg",
      "live_examples/live_01/frame_001.jpg",
      "live_examples/live_01/frame_002.jpg",
      "live_examples/live_01/frame_003.jpg",
      "live_examples/live_01/frame_004.jpg",
      "live_examples/live_01/frame_005.jpg",
      "live_examples/live_01/frame_006.jpg",
      "live_examples/live_01/frame_007.jpg",
      "live_examples/live_01/frame_008.jpg",
      "live_examples/live_01/frame_009.jpg",
      "live_examples/live_01/frame_010.jpg",
      "live_examples/live_01/frame_011.jpg",
      "live_examples/live_01/frame_012.jpg",
      "live_examples/live_01/frame_013.jpg",
      "live_examples/live_01/frame_014.jpg",
      "live_examples/live_01/frame_015.jpg",
      "live_examples/live_01/frame_016.jpg",
      "live_examples/live_01/frame_017.jpg",
      "live_examples/live_01/frame_018.jpg",
      "live_examples/live_01/frame_019.jpg",
      "live_examples/live_01/frame_020.jpg",
      "live_examples/live_01/frame_021.jpg",
      "live_examples/live_01/frame_022.jpg",
      "live_examples/live_01/frame_023.jpg",
      "live_examples/live_01/frame_024.jpg",
      "live_examples/live_01/frame_025.jpg",
      "live_examples/live_01/frame_026.jpg",
      "live_examples/live_01/frame_027.jpg",
      "live_examples/live_01/frame_028.jpg",
      "live_examples/live_01/frame_029.jpg",
      "live_examples/live_01/frame_030.jpg",
      "live_examples/live_01/frame_031.jpg",
      "live_examples/live_01/frame_032.jpg",
      "live_examples/live_01/frame_033.jpg",
      "live_examples/live_01/frame_034.jpg",
      "live_examples/live_01/frame_035.jpg",
      "live_examples/live_01/frame_036.jpg",
      "live_examples/live_01/frame_037.jpg",
      "live_examples/live_01/frame_038.jpg",
      "live_examples/live_01/frame_039.jpg",
      "live_examples/live_01/frame_040.jpg",
      "live_examples/live_01/frame_041.jpg",
      "live_examples/live_01/frame_042.jpg",
      "live_examples/live_01/frame_043.jpg",
      "live_examples/live_01/frame_044.jpg",
      "live_examples/live_01/frame_045.jpg",
      "live_examples/live_01/frame_046.jpg",
      "live_examples/live_01/frame_047.jpg",
      "live_examples/live_01/frame_048.jpg",
      "live_examples/live_01/frame_049.jpg",
      "live_examples/live_01/frame_050.jpg",
      "live_examples/live_01/frame_051.jpg",
      "live_examples/live_01/frame_052.jpg"
    ],
    "n_frames": 53,
    "test_idx": 1371
  },
  {
    "id": 2,
    "task": "cvs_assessment",
    "data_source": "Cholec80_CVS",
    "question": "You are an expert surgical video evaluator. This Cholec80-CVS recording is for assessing Strasberg\u2019s three Critical View of Safety criteria. Scoring the CVS requires evaluating three components: (1) Are there exactly two structures entering the gallbladder? (2) Is the cystic plate clearly seen? (3) Is the hepatocystic triangle free of tissue? Use a 0,1,2 scale for each.",
    "frames": [
      "live_examples/live_02/frame_000.jpg",
      "live_examples/live_02/frame_001.jpg",
      "live_examples/live_02/frame_002.jpg",
      "live_examples/live_02/frame_003.jpg",
      "live_examples/live_02/frame_004.jpg",
      "live_examples/live_02/frame_005.jpg",
      "live_examples/live_02/frame_006.jpg",
      "live_examples/live_02/frame_007.jpg",
      "live_examples/live_02/frame_008.jpg",
      "live_examples/live_02/frame_009.jpg",
      "live_examples/live_02/frame_010.jpg",
      "live_examples/live_02/frame_011.jpg",
      "live_examples/live_02/frame_012.jpg",
      "live_examples/live_02/frame_013.jpg",
      "live_examples/live_02/frame_014.jpg",
      "live_examples/live_02/frame_015.jpg",
      "live_examples/live_02/frame_016.jpg",
      "live_examples/live_02/frame_017.jpg",
      "live_examples/live_02/frame_018.jpg",
      "live_examples/live_02/frame_019.jpg",
      "live_examples/live_02/frame_020.jpg",
      "live_examples/live_02/frame_021.jpg",
      "live_examples/live_02/frame_022.jpg",
      "live_examples/live_02/frame_023.jpg",
      "live_examples/live_02/frame_024.jpg",
      "live_examples/live_02/frame_025.jpg",
      "live_examples/live_02/frame_026.jpg",
      "live_examples/live_02/frame_027.jpg",
      "live_examples/live_02/frame_028.jpg",
      "live_examples/live_02/frame_029.jpg",
      "live_examples/live_02/frame_030.jpg",
      "live_examples/live_02/frame_031.jpg",
      "live_examples/live_02/frame_032.jpg",
      "live_examples/live_02/frame_033.jpg",
      "live_examples/live_02/frame_034.jpg",
      "live_examples/live_02/frame_035.jpg",
      "live_examples/live_02/frame_036.jpg",
      "live_examples/live_02/frame_037.jpg",
      "live_examples/live_02/frame_038.jpg",
      "live_examples/live_02/frame_039.jpg",
      "live_examples/live_02/frame_040.jpg",
      "live_examples/live_02/frame_041.jpg",
      "live_examples/live_02/frame_042.jpg",
      "live_examples/live_02/frame_043.jpg",
      "live_examples/live_02/frame_044.jpg",
      "live_examples/live_02/frame_045.jpg",
      "live_examples/live_02/frame_046.jpg",
      "live_examples/live_02/frame_047.jpg",
      "live_examples/live_02/frame_048.jpg",
      "live_examples/live_02/frame_049.jpg",
      "live_examples/live_02/frame_050.jpg",
      "live_examples/live_02/frame_051.jpg",
      "live_examples/live_02/frame_052.jpg",
      "live_examples/live_02/frame_053.jpg",
      "live_examples/live_02/frame_054.jpg",
      "live_examples/live_02/frame_055.jpg",
      "live_examples/live_02/frame_056.jpg",
      "live_examples/live_02/frame_057.jpg",
      "live_examples/live_02/frame_058.jpg",
      "live_examples/live_02/frame_059.jpg"
    ],
    "n_frames": 60,
    "test_idx": 494
  }
]