On-Policy Distillation emerges as new AI training technique

Original: A new class of post-training method is emerging in 2026: On-Policy Distillation (OPD).

Source: x.com ↗