Datasets:
File size: 3,618 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 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 | {
"meta_info": {
"case_id": "case37_02",
"bench_subset": "NeuroBench-Multi",
"difficult": 2,
"original_dataset": "SEED-VIG"
},
"agent_input": {
"data_path": [
"data/multi/SEEDVIG_01_session-20151124_noon_2.edf"
],
"instruction": "Compare four 120-second intervals:\n\n- `interval_A`: [870, 990)\n- `interval_B`: [900, 1020)\n- `interval_C`: [2580, 2700)\n- `interval_D`: [4890, 5010)\n\nDivide each interval into consecutive 10-second epochs.\n\nFor each epoch, compute EEG slowing from `T7`, `T8`, `P1`, `P2`, `O1`, and `O2`. Average each band power across the six channels, then compute `log((theta + alpha) / (beta + gamma))`.\n\nFor `EOG-H` and `EOG-V`, average the `[0.1, 1.0)` Hz powers across the two channels and separately average the `[1.0, 10.0)` Hz powers, then compute `log(P_slow / P_fast)`.\n\nAverage the epoch-level EEG and EOG features within each interval and rank the four intervals from strongest to weakest evidence. Using all 48 paired epoch values, compute the Spearman correlation between EEG and EOG evidence.\n\nReport `eeg_drowsiness_order`, `eog_drowsiness_order`, and `eeg_eog_correlation`."
},
"eval_config": {
"parser_prompt": "### ROLE\nYou are a precise JSON Extraction Engine for EEG-EOG vigilance reports.\n\n### TASK\nExtract:\n- eeg_drowsiness_order: the four intervals ranked from strongest to weakest EEG drowsiness evidence.\n- eog_drowsiness_order: the four intervals ranked from strongest to weakest EOG drowsiness evidence.\n- eeg_eog_correlation: the Spearman correlation computed from all 48 paired epoch-level EEG and EOG values.\n\n### STRICT CONSTRAINTS\n1. Output ONLY one valid JSON object.\n2. Do not output Markdown, explanations, units, or extra fields.\n3. The keys must be exactly `eeg_drowsiness_order`, `eog_drowsiness_order`, and `eeg_eog_correlation`.\n4. Each order must be an array of exactly four labels in strongest-to-weakest order and must contain `A`, `B`, `C`, and `D` exactly once.\n5. Normalize interval names before output: `interval_A` and `A` both mean `A`; `interval_B` and `B` both mean `B`; `interval_C` and `C` both mean `C`; `interval_D` and `D` both mean `D`. In particular, `interval_D` and `D` are the same label.\n6. Do not output the `interval_` prefix in either order.\n7. `eeg_eog_correlation` must be a finite number in [-1, 1], or null if unavailable.\n\n### OUTPUT TEMPLATE\n{\"eeg_drowsiness_order\": [\"<A|B|C|D>\", \"<A|B|C|D>\", \"<A|B|C|D>\", \"<A|B|C|D>\"], \"eog_drowsiness_order\": [\"<A|B|C|D>\", \"<A|B|C|D>\", \"<A|B|C|D>\", \"<A|B|C|D>\"], \"eeg_eog_correlation\": <number|null>}",
"metrics": [
{
"metric_id": "eeg_drowsiness_order",
"type": "sequence_match_check",
"target_key": "eeg_drowsiness_order",
"weight": 35,
"params": {
"gt_value": [
"B",
"A",
"D",
"C"
],
"match_mode": "exact_order"
}
},
{
"metric_id": "eog_drowsiness_order",
"type": "sequence_match_check",
"target_key": "eog_drowsiness_order",
"weight": 35,
"params": {
"gt_value": [
"D",
"B",
"A",
"C"
],
"match_mode": "exact_order"
}
},
{
"metric_id": "eeg_eog_correlation",
"type": "numeric_check",
"target_key": "eeg_eog_correlation",
"weight": 30,
"params": {
"gt_value": 0.8527287537333695,
"tolerance": 0.08527287537333696
}
}
]
}
}
|