# Agentic On-Policy Distillation Implicitly Reshapes LLMs' ReAct Patterns

> Source: <https://aiflash.com/news/153544/>
> Published: 2026-11-18 05:02:14+00:00

**Recommendation:** After on-policy distillation, re-run ReAct interaction evals on out-of-domain benchmarks such as ALFWorld and τ^2-bench to catch shifted reasoning, action and observation patterns before deployment.
More in the playbook "Evals and testing for LLM apps": <https://wpnews.pro/playbooks/evals-testing>

On-policy distillation (OPD) transfers specialized capabilities to language-model agents, but its cross-domain effects on interaction remain unclear. We investigate how OPD reshapes ReAct interaction patterns across mathematics, ALFWorld, and τ^2-bench, focusing on reasoning, actions, and observatio
