A Unified Single-Pass Decision Model for EEG

EEG-Jev

Ask an EEG window a question in plain language. One shared model returns a typed answer with probabilities: a choice among candidates, the probability that a statement holds, or a score on an ordered scale.

Four trained questions play below on recorded EEG. Dashed cards are the next questions.

Recorded EEG one recorded EEG window
NoulVision Does this recording contain abnormal activity? → one probability
ChoiceVision Which artifact is this: eye blink, muscle, electrode pop, or none? → one label per window

Vision · next questions, not yet trainedAbnormal activityArtifact type

Continuous replay

One recording, end to end, with a saved answer for every window.

Question

Choice answer · a probability for each of the five sleep stages in this 30-s epoch. Noul answer · Model P(seizure overlap): the probability that an annotated seizure overlaps this 5-s window. Choice answer · a probability for each motor-imagery task offered in the question. Score answer · a probability for each rating from 0 to 10; the expected value is read from it.

4×
Model’s answer saved output
Signal cue rule-based measurement
Dataset label from the dataset

Window by window ↓

Recorded replaySaved model outputData & methods

02 · Window by window

Explore individual EEG windows

Curated examples from other recordings, with EEG traces, dataset annotations and saved model answers.

01 / 10

1EEG window
—
  • Reference label dataset
  • Signal view computed in your browser
  • Saved model answer recorded output
Prediction window30 sone saved answer per window
Playback≈4 s per windowdrawing speed, not inference time
Displayed rangethis window onlySingle windows · full recording in the replay
EEG bipolar

EEG: voltage (µV) over time. The axis is seconds within this window.

Signal analysis Measurements of this window, computed in your browser · not the model’s reasons

Point at a picture to read a value.

Spectrogram which rhythms are strong, and when
  • Acrosstime within this window, on the same axis as the traces above.
  • Upfrequency, from slow to fast (Hz).
  • Colourpower spectral density in dB; one scale for every example of this task.
weakerstronger
Measurement

    Loading recorded material…

    How it works

    EEG and a question in, a typed answer out

    A frozen REVE-L encoder turns the EEG window into features, and a trainable MLP connector maps them into the embedding space of Qwen3.5-2B, whose frozen base weights are adapted by LoRA. The EEG, the question and the answer specification pass through the model once. At a learned readout marker, the readout selected by the answer type returns probabilities; no text is generated.

    The research

    EEG-Jev: A Unified Single-Pass Decision Model for EEG

    EEG-Jev is evaluated on selected NeuralBench datasets, not on the whole benchmark. Data & methods