{ "@context": { "@language": "en", "@vocab": "https://schema.org/", "citeAs": "cr:citeAs", "column": "cr:column", "conformsTo": "dct:conformsTo", "cr": "http://mlcommons.org/croissant/", "rai": "http://mlcommons.org/croissant/RAI/", "data": { "@id": "cr:data", "@type": "@json" }, "dataType": { "@id": "cr:dataType", "@type": "@vocab" }, "dct": "http://purl.org/dc/terms/", "examples": { "@id": "cr:examples", "@type": "@json" }, "extract": "cr:extract", "field": "cr:field", "fileProperty": "cr:fileProperty", "fileObject": "cr:fileObject", "fileSet": "cr:fileSet", "format": "cr:format", "includes": "cr:includes", "isLiveDataset": "cr:isLiveDataset", "jsonPath": "cr:jsonPath", "key": "cr:key", "md5": "cr:md5", "parentField": "cr:parentField", "path": "cr:path", "recordSet": "cr:recordSet", "references": "cr:references", "regex": "cr:regex", "repeated": "cr:repeated", "replace": "cr:replace", "sc": "https://schema.org/", "separator": "cr:separator", "source": "cr:source", "subField": "cr:subField", "transform": "cr:transform" }, "@type": "sc:Dataset", "conformsTo": "http://mlcommons.org/croissant/1.0", "name": "Synapse", "description": "Synchronized Neural and Auditory Processing for Sensory Evaluation (Synapse). A 16-channel cEEGrid around-the-ear EEG dataset recorded during passive auditory stimulation from 23 adult subjects (16 EXP: sound-sensitivity disorder; 7 CTRL). Released as the validation testbed for auditing EEG foundation-model transfer to passive wearable clinical settings.", "citeAs": "Anonymous Authors. Implicit Choices, Explicit Failures: Auditing EEG Foundation-Model Transfer to Passive Wearable Clinical Settings. NeurIPS 2026 Evaluations and Datasets Track.", "license": "https://creativecommons.org/licenses/by/4.0/", "url": "https://huggingface.co/datasets/synapse-neurips2026/synapse", "version": "1.0.0", "datePublished": "2026-05-01", "isLiveDataset": false, "keywords": [ "EEG", "ear-EEG", "cEEGrid", "wearable EEG", "hyperacusis", "tinnitus", "misophonia", "sound sensitivity", "auditory", "passive paradigm", "clinical", "foundation model", "transfer learning" ], "creator": { "@type": "sc:Organization", "name": "Anonymous Institution (to be updated at camera-ready)" }, "rai:dataCollectionType": "Controlled human subjects experiment with IRB approval. Participants wore 16-channel cEEGrid ear-EEG in a sound-attenuated booth and passively listened during three tasks: LET (listening effort, speech-in-noise), HLT (hearing loudness test), AST (aversive sound test). No active cognitive task was required.", "rai:dataBiases": "Cohort imbalance: EXP averages 41.7 years (range 19-73), CTRL averages 25.3 years (range 21-35). CTRL is 100% female; EXP is 11M/5F. Age- and sex-related EEG spectral differences may partly drive absolute AUROC values and should not be interpreted as disorder-specific at n=23.", "rai:personalSensitiveInformation": "Clinical disorder diagnoses (hyperacusis, tinnitus, misophonia) and questionnaire scores (HQ, THI, MQ, GAD-7, Iowa) are sensitive health information. Subject IDs are fully anonymized; no directly identifying information is included. All data are released under CC-BY-4.0.", "rai:annotationsPerItem": "Per subject: (1) binary EXP/CTRL label; (2) continuous clinical scores: HQ, THI, MQ, GAD-7, Iowa; (3) per-channel boolean QC mask (16 channels); (4) per-task trial-level EEG epochs.", "rai:labelingPersonas": "Clinical labels assigned by trained clinicians using validated questionnaire thresholds (THI>=18, HQ>=28, MQ>=5). Per-channel QC labels assigned by automated gates reviewed by a trained research assistant.", "rai:dataLimitations": "n=23 limits statistical power; paired-bootstrap CIs are wide. Cohort imbalance in age (EXP mean 41.7y vs CTRL 25.3y) and sex (CTRL 100% female) may inflate absolute AUROC values. Single recruitment site and single clinical domain (sound-sensitivity disorder). Frozen-feature evaluation only; fine-tuning results not included.", "rai:dataUseCases": "Intended use: validation testbed for auditing EEG foundation-model transfer to passive wearable clinical settings. Not intended for: clinical-grade screening or diagnosis at n=23; population-level disorder generalization; standalone deployment without expert supervision.", "rai:dataSocialImpact": "Positive: enables rigorous auditing of EEG foundation models before clinical deployment, reducing risk of deploying poorly transferring models in wearable health settings. Negative: absolute AUROC values may be misinterpreted as clinical-grade performance; cohort imbalance may limit generalizability to broader populations.", "rai:hasSyntheticData": false, "distribution": [ { "@type": "cr:FileObject", "@id": "eeg-pkl", "name": "synapse_125hz_pre1s_post7s.pkl", "description": "Main EEG data file (Python pickle). Dict with keys 'records' (92 subject-task dicts), 'meta' (dataset parameters), 'subject_quality' (per-subject QC). Each record: subject, group, task, X (n_trials x 16 x 1000 float32), channel_mask (16 bool), n_trials.", "contentUrl": "https://huggingface.co/datasets/synapse-neurips2026/synapse/resolve/main/synapse_125hz_pre1s_post7s.pkl", "encodingFormat": "application/octet-stream", "sha256": "0416cc7acc80aad37ad0aaae9bf5318acc778b35aa51b84fd7e560abd21ee8d6" }, { "@type": "cr:FileObject", "@id": "results-zip", "name": "results_96cell.zip", "description": "Per-cell subject-level probability vectors for all 96 cells of the stress-test grid. Each JSON: predicted P(EXP) for 23 LOSO subjects, seed-level probabilities, AUROC, configuration metadata.", "contentUrl": "https://huggingface.co/datasets/synapse-neurips2026/synapse/resolve/main/results_96cell.zip", "encodingFormat": "application/zip", "sha256": "0aeee311ecc7bd14409d5c384a7d15f0adf9013a1afdc6cdd7231e5bffb90bec" }, { "@type": "cr:FileObject", "@id": "code-zip", "name": "code.zip", "description": "Full reproduction code: feature extraction for 8 foundation models, 4-class probe suite, LOSO evaluation, bootstrap statistics, figure generation.", "contentUrl": "https://huggingface.co/datasets/synapse-neurips2026/synapse/resolve/main/code.zip", "encodingFormat": "application/zip", "sha256": "15e9f95931670f0dff592ea884fa649f8489cb6ca569a12b839cff79a2f8d96e" } ], "recordSet": [ { "@type": "cr:RecordSet", "@id": "eeg-records", "name": "Subject-task EEG records", "description": "92 records (23 subjects x 4 tasks). Source: synapse_125hz_pre1s_post7s.pkl, key 'records'.", "field": [ { "@type": "cr:Field", "@id": "eeg-records/subject", "name": "subject", "description": "Anonymized subject ID (e.g., EXP01, CTRL06). 16 EXP and 7 CTRL subjects.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "subject" } } }, { "@type": "cr:Field", "@id": "eeg-records/group", "name": "group", "description": "Binary group label: EXP (sound-sensitivity disorder) or CTRL (healthy control).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "group" } } }, { "@type": "cr:Field", "@id": "eeg-records/task", "name": "task", "description": "Task ID: 'let' (listening effort), 'hlt' (hearing loudness), 'ast' (aversive sound), 'pmt' (passive monitoring).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "task" } } }, { "@type": "cr:Field", "@id": "eeg-records/n_trials", "name": "n_trials", "description": "Number of valid trials for this subject-task pair (range 4-35).", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "n_trials" } } }, { "@type": "cr:Field", "@id": "eeg-records/X_shape", "name": "X_shape", "description": "EEG epoch tensor shape: (n_trials, 16, 1000). 16 cEEGrid channels (L01-L10, R01-R10, positions 3 and 6 absent). 1000 samples at 125 Hz covering [-1, +7] s peri-stimulus. dtype=float32. No normalization applied.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "X" } } }, { "@type": "cr:Field", "@id": "eeg-records/channel_mask", "name": "channel_mask", "description": "Boolean QC mask shape (16,). True = channel passed impedance/variance/correlation gates.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "channel_mask" } } } ] }, { "@type": "cr:RecordSet", "@id": "clinical-scores", "name": "Subject-level clinical labels and scores", "description": "23 subject-level records with binary group labels and continuous clinical questionnaire scores (HQ, THI, MQ, GAD-7, Iowa). Released under data-use agreement.", "field": [ { "@type": "cr:Field", "@id": "clinical-scores/subject", "name": "subject", "description": "Anonymized subject ID, matching eeg-records.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "subject" } } }, { "@type": "cr:Field", "@id": "clinical-scores/group", "name": "group", "description": "Binary group label: EXP or CTRL.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "group" } } }, { "@type": "cr:Field", "@id": "clinical-scores/HQ", "name": "HQ", "description": "Hyperacusis Questionnaire total score (0-100). Clinical threshold >= 28.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "HQ" } } }, { "@type": "cr:Field", "@id": "clinical-scores/THI", "name": "THI", "description": "Tinnitus Handicap Inventory total score (0-100). Clinical threshold >= 18.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "THI" } } }, { "@type": "cr:Field", "@id": "clinical-scores/MQ", "name": "MQ", "description": "Misophonia Questionnaire total score. Clinical threshold >= 5.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "MQ" } } }, { "@type": "cr:Field", "@id": "clinical-scores/GAD7", "name": "GAD-7", "description": "Generalized Anxiety Disorder-7 total score (0-21). Clinical threshold >= 10.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "GAD7" } } }, { "@type": "cr:Field", "@id": "clinical-scores/Iowa", "name": "Iowa", "description": "Iowa Tinnitus Activities Questionnaire subscale scores.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "eeg-pkl" }, "extract": { "column": "Iowa" } } } ] } ] }