Nvidia adopts multi-teacher distillation for AI model training

Original: Nvidia joined the multi-teacher, on-policy distillation (MODP) gang! Is industry standard post-training right now.

Source: x.com ↗