TY  - RPRT
TI  - Is Per-Agent Policy Composition Safe? Rethinking Successor-Feature Transfer in Cooperative Multi-Agent Reinforcement Learning
AU  - Zijian Zhao
AU  - Sen Li
PY  - 2026
UR  - https://arxiv.org/abs/2608.11658
ID  - 2608.11658
ER  -