| {
|
| "meta_info": {
|
| "case_id": "SEED-V-02.cnt",
|
| "bench_subset": "NeuroBench-Core",
|
| "difficult": 2.0,
|
| "original_dataset": "SEED-V"
|
| },
|
| "agent_input": {
|
| "data_path": "data/core/SEED-V-02.cnt",
|
| "instruction": "Please select standard EEG channels. For each delta, theta, alpha, and beta band, apply a zero-phase FIR bandpass filter to the selected continuous signals, then use the first 120 seconds of the filtered data. Extract instantaneous phase with the Hilbert transform, compute PLV for every unique channel pair, and define the global mean PLV as the arithmetic mean across those pairs. Report the band with the highest global mean PLV and briefly explain the result."
|
| },
|
| "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 five outputs from the agent report:\n1) delta global mean PLV\n2) theta global mean PLV\n3) alpha global mean PLV\n4) beta global mean PLV\n5) dominant synchronization band\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 must be EXACTLY \"delta_global_mean_plv\", \"theta_global_mean_plv\", \"alpha_global_mean_plv\", \"beta_global_mean_plv\", and \"dominant_synchronization_band\".\n5. The four PLV values must be float or null.\n6. \"dominant_synchronization_band\" must be one of \"delta\", \"theta\", \"alpha\", \"beta\", or null.\n\n### OUTPUT TEMPLATE\n{\"delta_global_mean_plv\": <float|null>, \"theta_global_mean_plv\": <float|null>, \"alpha_global_mean_plv\": <float|null>, \"beta_global_mean_plv\": <float|null>, \"dominant_synchronization_band\": <string|null>}",
|
| "metrics": [
|
| {
|
| "metric_id": "delta_global_mean_plv_accuracy",
|
| "type": "numeric_check",
|
| "target_key": "delta_global_mean_plv",
|
| "weight": 15,
|
| "params": {
|
| "gt_value": 0.31769439412145845,
|
| "tolerance": 0.015884719706072924
|
| }
|
| },
|
| {
|
| "metric_id": "theta_global_mean_plv_accuracy",
|
| "type": "numeric_check",
|
| "target_key": "theta_global_mean_plv",
|
| "weight": 15,
|
| "params": {
|
| "gt_value": 0.32533324405965947,
|
| "tolerance": 0.016266662202982975
|
| }
|
| },
|
| {
|
| "metric_id": "alpha_global_mean_plv_accuracy",
|
| "type": "numeric_check",
|
| "target_key": "alpha_global_mean_plv",
|
| "weight": 15,
|
| "params": {
|
| "gt_value": 0.31677480733100466,
|
| "tolerance": 0.015838740366550235
|
| }
|
| },
|
| {
|
| "metric_id": "beta_global_mean_plv_accuracy",
|
| "type": "numeric_check",
|
| "target_key": "beta_global_mean_plv",
|
| "weight": 15,
|
| "params": {
|
| "gt_value": 0.2656337040205575,
|
| "tolerance": 0.013281685201027876
|
| }
|
| },
|
| {
|
| "metric_id": "dominant_band_explanation_consistency",
|
| "type": "semantic_consistency",
|
| "target_key": "full_text",
|
| "weight": 40,
|
| "params": {
|
| "judge_prompt": "Judge whether the final dominant synchronization band and its explanation are consistent with the true ranking of delta/theta/alpha/beta global mean PLV values. The true dominant synchronization band is theta. Mark the response as correct only if it selects the true highest-PLV band and the explanation clearly supports that band as the strongest synchronization band. Output only a valid JSON object with exactly one key named status and a boolean value True or False."
|
| }
|
| }
|
| ]
|
| }
|
| }
|
|
|
