Datasets:
| { | |
| "meta_info": { | |
| "case_id": "case02_09", | |
| "bench_subset": "NeuroBench-Multi", | |
| "difficult": 1, | |
| "original_dataset": "SEED-VIG" | |
| }, | |
| "agent_input": { | |
| "data_path": "data/multi/SEEDVIG_04_session-20151105_noon.edf", | |
| "instruction": "Count analyzable physiological channels for modality EOG and modality fNIRS in the provided recording. Do not count markers, status, labels, timestamps, metadata, or trial fields. Return only modality_1_channel_count and modality_2_channel_count." | |
| }, | |
| "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 the following final answer fields from the agent's report into one JSON object:\n- modality_1_channel_count: Modality 1 channel count.\n- modality_2_channel_count: Modality 2 channel count.\n\n### STRICT CONSTRAINTS (MANDATORY)\n1. Output ONLY a valid JSON object.\n2. DO NOT include Markdown code blocks.\n3. DO NOT include conversational text, explanations, or extra fields.\n4. The JSON keys must be EXACTLY the keys listed in the OUTPUT TEMPLATE.\n5. Preserve the declared value type for every key.\n6. If a requested value is missing or cannot be found, return null for that key.\n\n### OUTPUT TEMPLATE\n{\"modality_1_channel_count\": <integer>, \"modality_2_channel_count\": <integer>}", | |
| "metrics": [ | |
| { | |
| "metric_id": "modality_1_channel_count_accuracy", | |
| "type": "numeric_check", | |
| "target_key": "modality_1_channel_count", | |
| "weight": 50, | |
| "params": { | |
| "gt_value": 2, | |
| "tolerance": 0.0 | |
| } | |
| }, | |
| { | |
| "metric_id": "modality_2_channel_count_accuracy", | |
| "type": "numeric_check", | |
| "target_key": "modality_2_channel_count", | |
| "weight": 50, | |
| "params": { | |
| "gt_value": 0, | |
| "tolerance": 0.0 | |
| } | |
| } | |
| ] | |
| } | |
| } | |