Datasets:
File size: 2,034 Bytes
752e201 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 | {
"meta_info": {
"case_id": "case29_02",
"bench_subset": "NeuroBench-Multi",
"difficult": 1,
"original_dataset": "SEED-VIG"
},
"agent_input": {
"data_path": "data/multi/SEEDVIG_02_session-20151106_noon.edf",
"instruction": "Analyze [824, 1004) seconds using six consecutive 30-second windows.\nFor each window, summarize EEG drowsiness evidence as the median across EEG channels of log(theta / (alpha + beta)), using theta [4, 8), alpha [8, 14), and beta [14, 31) Hz. Summarize EOG slow-eye-movement evidence as the mean [0.1, 1.0) Hz power across EOG-H and EOG-V.\nAverage windows 1–2 as early, 3–4 as middle, and 5–6 as late. For each modality, report the phase with the strongest evidence.\nReport only eeg_peak_phase and eog_peak_phase."
},
"eval_config": {
"parser_prompt": "### ROLE\nYou are a precise JSON Extraction Engine for multimodal drowsiness reports. Your sole task is to convert the agent's report into one JSON object.\n\n### TASK\nExtract:\n- eeg_peak_phase\n- eog_peak_phase\n\n### STRICT CONSTRAINTS (MANDATORY)\n1. Output ONLY a valid JSON object.\n2. DO NOT include Markdown, explanations, or extra fields.\n3. The JSON keys must be EXACTLY \"eeg_peak_phase\" and \"eog_peak_phase\".\n4. Each value must be exactly \"early\", \"middle\", or \"late\", or null if missing or unclear.\n5. Extract the phase with the strongest evidence for each modality, not intermediate window results.\n\n### OUTPUT TEMPLATE\n{\"eeg_peak_phase\": \"<early|middle|late|null>\", \"eog_peak_phase\": \"<early|middle|late|null>\"}",
"metrics": [
{
"metric_id": "eeg_peak_phase",
"type": "categorical_check",
"target_key": "eeg_peak_phase",
"weight": 50,
"params": {
"gt_value": "late"
}
},
{
"metric_id": "eog_peak_phase",
"type": "categorical_check",
"target_key": "eog_peak_phase",
"weight": 50,
"params": {
"gt_value": "early"
}
}
]
}
}
|