JD Proposes ROAD MOPD: Cross-Teacher Guarding for Multi-Teacher Distillation

JD, together with Nanjing University, published a paper proposing ROAD MOPD, which introduces cross-teacher guarding and dual-teacher consistency gating to suppress gradient conflicts in multi-teacher distillation, improving upon MOPD.

2026-10-22 ~ 2026-10-22 · 2 related posts