Feed
Decoding silent reading from non-invasive EEG
49 hours of one person's EEG while reading lets a model pick out individual words.
7.5
pp
Word-level retrieval gain above chance
Decoding works even where topic tracking is impossible
How it works
No inner-speech labels
Nobody can record what a person silently thinks with exact timing, so inner-speech decoders lack training data.
Silent reading as proxy
One participant read ~240,000 words in random fonts while dry EEG was recorded; a contrastive model matched EEG to Llama embeddings.
Words decodable, data-limited
Every fit beat chance, rare words decoded too, and accuracy kept rising log-linearly with more data.
From the paper