Branching On-Policy Distillation
On-policy distillation (OPD) provides teacher supervision on student-generated trajectories, but a single trajectory realizes only one continuation at each decision, limiting supervision coverage under a finite sampling budget. We propose BranchOPD (Branching On-Policy Distillation), which selects informative decision...
Dong-Yang Chen, Shudong Liu, Shu-Bo Li et al.
· 0 citations