Multi-Token Prediction Done Right
Paper • 2610.38331 • Published • 9
Pre-trained models from the paper "Multi-Token Prediction Done Right"
Note 167m — next-token baseline
Note 167m — standard MTP
Note 167m — Soft-MTP
Note 332m — next-token baseline
Note 332m — standard MTP
Note 332m — Soft-MTP
Note 689m — next-token baseline
Note 689m — standard MTP
Note 689m — Soft-MTP
Note 1p2b — next-token baseline
Note 1p2b — standard MTP
Note 1p2b — Soft-MTP