MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation ▲ 5 💬 1 2026-10-27 arXiv HF 原文 摘要 （无摘要）