Road MOPD: resolving multi-teacher distillation conflicts, +5.6–8.8 pts on medical tasks
jiqizhixin · x · 2026-11-20
A Domai and Nanjing University paper asks what happens when specialist teachers train the same student: even with one teacher per domain, their updates still act on shared student parameters, creating gradient conflicts. Road MOPD lets an auxiliary teacher score tokens in the student's own rollouts, with agreement-modulated weighting, plus gradient projection that removes components opposing the auxiliary objective while retaining minimum gradient norm. In high-conflict math and medical settings this improves over vanilla multi-teacher on-policy distillation by 5.6–8.8 percentage points (specific domain pairs and setups). The repo says code release is pending approval.
More from Research
- Testing browser agents on the live web is unsafe, so this team built a hermetic synthetic "Mirror Dimension" — laparisa · 2026-11-20
- Roboticist argues full joint space is the wrong abstraction for high-level agents — KyleMorgenstein · 2026-11-20
- Do language models have explicit goal encodings? New interpretability paper takes first step — benno_krojer · 2026-11-20
- Regularized Score Centering fixes advantage estimation in agentic RL, 9.9× on BabyAI — aviral_kumar2 · 2026-11-20
- DeepMind Adds Memory to Recursive Self-Improvement, Boosting Scores 68→85 — Dr_Singularity · 2026-11-20
- LLM Malaria Drug Discovery Paper Accepted at NeurIPS Africa in AI Workshop — madeofAjala · 2026-11-20