Skip to main content
Search

# Multi-Teacher On-Policy Distillation

1 related articles

Subscribe to updates
  1. arXiv

    Agentic on-policy distillation rewrites ReAct habits: an ALFWorld teacher drops customer-service accuracy from 41.5% to 6.6%

    On-policy distillation transfers a domain teacher's capability to language-model agents, but across domains it also rewrites how they organize reasoning, actions and observation gathering. Comparing students distilled from different sources on mathematics, ALFWorld and customer-service tool tasks, code-interpreter supervision raised customer-service accuracy from 41.5% to 51.0%, while ALFWorld supervision cut it to 6.6% and brought many tool calls issued before the user replied. Using only source-domain trajectories, the authors then selected supervised positions by the teacher's log-probability change before and after domain RL; keeping the top 95% raised customer-service accuracy from 6.6% to 68.9% at some cost to source-domain performance.

You’re all caught up