Datasets:
| { | |
| "meta_info": { | |
| "case_id": "Q01", | |
| "bench_subset": "NeuroBench-Core", | |
| "difficult": 1.0, | |
| "original_dataset": "Core_QA", | |
| "references": "R1,R2" | |
| }, | |
| "agent_input": { | |
| "data_path": "", | |
| "instruction": "This is a standalone EEG/BCI knowledge multiple-choice question. No EEG data file is needed. Please answer by selecting exactly one option: A, B, C, or D.\n\nQuestion: What type of neural activity is scalp EEG thought to reflect most directly?\nA. The synchronous summation of single-neuron axonal action potentials at the scalp.\nB. Field potentials generated by postsynaptic currents in large populations of similarly oriented cortical pyramidal cells.\nC. Deep thalamic action potentials amplified by the skull.\nD. Fast projections of cerebral blood-flow and oxygen-metabolism changes." | |
| }, | |
| "eval_config": { | |
| "parser_prompt": "### ROLE\nYou are a precise JSON Extraction Engine for multiple-choice EEG/BCI QA responses. Your sole task is to extract the final selected answer option.\n\n### TASK\nExtract the final selected option from the agent report.\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, explanations, or units.\n4. The key must be EXACTLY \"selected_option\".\n5. The value must be one of \"A\", \"B\", \"C\", \"D\", or null.\n6. If multiple options are mentioned, extract only the final answer option.\n7. If the selected option is missing or unclear, return {\"selected_option\": null}.\n\n### OUTPUT TEMPLATE\n{\"selected_option\": <string|null>}", | |
| "metrics": [ | |
| { | |
| "metric_id": "qa_option_accuracy", | |
| "type": "categorical_check", | |
| "target_key": "selected_option", | |
| "weight": 100, | |
| "params": { | |
| "gt_value": "B" | |
| } | |
| } | |
| ] | |
| } | |
| } |