#artificial intelligence
Jun 2026
Staying In Character: Perspective-Bounded Memory For Book-Based Role-Playing Agents
RverieMem is proposed, a three-layer memory architecture for book-based character agents that improves both boundary fidelity and character-grounded narrative generation and achieves a mean win rate of 78.86% against BookWorld's five-dimension pairwise narrative protocol.
Xu-Shuo Tang, Junhe Zhang, Zi-Han Yang et al.
· arXiv.org · 3 citations