Toward a realistic model of speech processing in the brain with\n self-supervised learning
Bibliographic record
Abstract
Several deep neural networks have recently been shown to generate activations\nsimilar to those of the brain in response to the same input. These algorithms,\nhowever, remain largely implausible: they require (1) extraordinarily large\namounts of data, (2) unobtainable supervised labels, (3) textual rather than\nraw sensory input, and / or (4) implausibly large memory (e.g. thousands of\ncontextual words). These elements highlight the need to identify algorithms\nthat, under these limitations, would suffice to account for both behavioral and\nbrain responses. Focusing on the issue of speech processing, we here\nhypothesize that self-supervised algorithms trained on the raw waveform\nconstitute a promising candidate. Specifically, we compare a recent\nself-supervised architecture, Wav2Vec 2.0, to the brain activity of 412\nEnglish, French, and Mandarin individuals recorded with functional Magnetic\nResonance Imaging (fMRI), while they listened to ~1h of audio books. Our\nresults are four-fold. First, we show that this algorithm learns brain-like\nrepresentations with as little as 600 hours of unlabelled speech -- a quantity\ncomparable to what infants can be exposed to during language acquisition.\nSecond, its functional hierarchy aligns with the cortical hierarchy of speech\nprocessing. Third, different training regimes reveal a functional\nspecialization akin to the cortex: Wav2Vec 2.0 learns sound-generic,\nspeech-specific and language-specific representations similar to those of the\nprefrontal and temporal cortices. Fourth, we confirm the similarity of this\nspecialization with the behavior of 386 additional participants. These\nelements, resulting from the largest neuroimaging benchmark to date, show how\nself-supervised learning can account for a rich organization of speech\nprocessing in the brain, and thus delineate a path to identify the laws of\nlanguage acquisition which shape the human brain.\n
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".