Fig. 4: Cross-modal generalization: MEG identification accuracy using the fMRI-derived brain-aligned semantic vectors.
From: Brain-aligning of semantic vectors improves neural decoding of visual stimuli
Violin plots showing zero-shot identification accuracy for visual stimulus decoding in MEG data using brain-aligned semantic vectors derived from the fMRI V4 region. a CLIP-based and b GloVe-based semantic vectors with optimal α parameters determined from the fMRI results. The individual data points from 3 subjects are overlaid on violin plots. The white circles indicate shuffled control data, demonstrating chance-level performance.
