OPMD: On-Policy Mechanism Distillation Beyond Output Supervision
On-policy distillation (OPD) improves language-model by supervising a student on prefixes sampled from its own policy, but its supervision remains confined to the output space. Consequently, OPD transfers what a teacher predicts without supervising the intermediate computation that produces those predictions. Simply di...
Ying-Dong Shi, Zhi-Hong Li, Chang-Ming Li et al.
· 1 citation