Datasets:
File size: 3,922 Bytes
d846fa6 | 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 61 62 63 64 65 66 | {
"meta_info": {
"case_id": "Mumtaz2016_02.edf",
"bench_subset": "NeuroBench-Core",
"difficult": 1.5,
"original_dataset": "Mumtaz2016"
},
"agent_input": {
"data_path": "data/core/Mumtaz2016_02.edf",
"instruction": "Please select EEG channels, downsample to 200Hz, apply a 0.5-45Hz FIR bandpass filter, compute channel-wise kurtosis and skewness, then list four channel names in descending order by kurtosis and four channel names in descending order by skewness."
},
"eval_config": {
"parser_prompt": "### ROLE\nYou are a precise JSON Extraction Engine for neurophysiological data. Your sole task is to convert natural language reports into structured JSON data.\n\n### TASK\nExtract two ranked channel sequences from the agent report:\n1) Top-4 channels by kurtosis (descending)\n2) Top-4 channels by skewness (descending)\n\n### STRICT CONSTRAINTS (MANDATORY)\n1. Output ONLY a valid JSON object.\n2. DO NOT include Markdown code blocks.\n3. DO NOT include any conversational text or explanations.\n4. Keys in JSON must be EXACTLY \"kurtosis_top4_channels\" and \"skewness_top4_channels\".\n5. Extract the two final ranked lists, not channel inventories or intermediate values. Each value must be an array of exactly 4 strings.\n6. Normalize EACH reported channel independently: trim whitespace, remove one leading acquisition prefix EEG followed by a space, hyphen, or underscore, then uppercase the label.\n7. Remove one terminal reference suffix if present: -A1, -A2, -LE, -REF, -M1, -M2, or -AVG.\n8. Convert legacy aliases after removing the prefix and suffix: T3->T7, T4->T8, T5->P7, T6->P8.\n9. Each output label must contain only uppercase A-Z letters and digits. Examples: O2-A1 -> O2; EEG Fp1-LE -> FP1; EEG T4 -> T8; Cz -> CZ.\n10. Preserve the descending rank order within each list. Do not sort either array alphabetically and do not substitute different channels.\n11. If one final list does not contain exactly 4 identifiable channel labels, return null only for that value; do not guess.\n\n### OUTPUT TEMPLATE\n{\"kurtosis_top4_channels\": [\"CH1\",\"CH2\",\"CH3\",\"CH4\"], \"skewness_top4_channels\": [\"CH1\",\"CH2\",\"CH3\",\"CH4\"]}",
"metrics": [
{
"metric_id": "kurtosis_top4_sequence",
"type": "sequence_match_check",
"target_key": "kurtosis_top4_channels",
"weight": 50,
"params": {
"gt_value": [
"T7",
"T8",
"P7",
"F8"
],
"match_mode": "weighted_partial_order",
"top_k": 4,
"position_weights": [
1.0,
0.8,
0.6,
0.4
],
"min_overlap": 2,
"allow_order_slip": 1
}
},
{
"metric_id": "skewness_top4_sequence",
"type": "sequence_match_check",
"target_key": "skewness_top4_channels",
"weight": 50,
"params": {
"gt_value": [
"P7",
"P3",
"FZ",
"O2"
],
"match_mode": "weighted_partial_order",
"top_k": 4,
"position_weights": [
1.0,
0.8,
0.6,
0.4
],
"min_overlap": 2,
"allow_order_slip": 1
}
}
]
}
}
|