Fine-tuning language encoding models on slow fMRI improves prediction for fast ECoG

In three linesResearchers use functional MRI (fMRI) to improve encoding models trained on ECoG (electrocorticography). By fine-tuning spoken language representations on fMRI, they achieve better ECoG predictions despite 100× lower temporal resolution. Performance improves with more fMRI training data.

Summary generated by Claude — human-verified