{
  "meta_info": {
    "case_id": "case13_35",
    "bench_subset": "NeuroBench-Multi",
    "difficult": 1.5,
    "original_dataset": "SHHS1"
  },
  "agent_input": {
    "data_path": "data/multi/SHHS1_01.edf",
    "instruction": "Analyze EEG channel `EEG(sec)` and GSR channel `GSR1` over [0, 600).\n\nDivide both streams into aligned 60-second windows with a 30-second stride, and pair the windows by index.\n\nFor each window, compute `delta_power` from the EEG signal as [0.5, 4) Hz band power. Compute `scl_mean` from the GSR signal as mean skin-conductance level.\n\nKeep windows with finite values for both features. If at least 6 valid pairs remain and both feature sequences have non-zero variance, compute their Pearson correlation. Otherwise, the sign cannot be determined.\n\nReport correlation_sign as `positive` when the correlation is greater than zero and `negative` when it is less than zero. Report null when the correlation cannot be determined or is exactly zero.\n\nReport only correlation_sign."
  },
  "eval_config": {
    "parser_prompt": "### ROLE\nYou are a strict JSON formatting engine.\n\n### TASK\nThis case is evaluated from the agent's complete response by a semantic judge.\nReturn an empty JSON object.\n\n### STRICT CONSTRAINTS\n1. Output ONLY a valid empty JSON object.\n2. Do not output Markdown, explanations, or additional fields.\n\n### OUTPUT TEMPLATE\n{}",
    "metrics": [
      {
        "metric_id": "missing_modality_channel_handling",
        "type": "semantic_consistency",
        "target_key": "full_text",
        "weight": 100,
        "params": {
          "judge_prompt": "Judge whether the response correctly handles an unavailable modality or channel in the supplied recording.\n\nThe requested `GSR` modality and/or channel(s) `GSR1` are not present in the file specified by `data_path`, so the requested window-level feature sequence and `correlation_value` cannot be computed.\n\nA correct response must:\n1. Clearly state that the requested modality or channel is unavailable in the supplied recording.\n2. State or clearly imply that the requested correlation cannot therefore be computed.\n3. Avoid fabricating, estimating, or inferring the missing physiological signal.\n4. Avoid reporting a numerical `correlation_value`, including `0.0`, as though it were a computed result.\n5. Base the conclusion on data availability rather than physiological interpretation, statistical significance, or data quality speculation.\n\nThe exact wording and level of detail may vary. Return {\"status\": true} only when all applicable requirements are satisfied; otherwise return {\"status\": false}. Output exactly one JSON object with the single boolean key `status`, without explanations, Markdown, scores, or additional fields."
        }
      }
    ]
  }
}