Assistant Professor, International Higher School for Advanced Studies Trieste
1 paper at NeurIPS 2025
RoMAE is a single, drop-in masked autoencoder that utilizes continuous (axial) RoPE to excel in interpolation and representation learning across modalities.
