Source note2 min read
Staying In Character: Perspective-Bounded Memory For Book-Based Role-Playing Agents
Source
arXiv
arxiv.org/abs/2606.25632v1
Source
- Source name: arXiv
- Source kind: paper
- Authors / organization: Xushuo Tang, Junhe Zhang, Zihan Yang, Yifu Tang, Sichao Li, Longbin Lai, Zhengyi Yang
- Venue: arXiv: cs.CL
- Published: 2026-06-24
- Collected: 2026-06-25
- URL: http://arxiv.org/abs/2606.25632v1
- PDF: https://arxiv.org/pdf/2606.25632v1
- DOI: UNKNOWN
Plain-Language Brief
Researchers propose a three-layer memory architecture named REVERIEMEM for book-based role-playing agents to address the issues of Factual Overreach and Stylistic Monotony in current LLM role-playing systems. The evaluation of this method shows improved performance over previous methods.
What the Document Says
- The paper discusses a proposed memory architecture called REVERIEMEM for character agents in book-based role-playing.
- REVERIEMEM has three layers: episodic, semantic, and personality.
- The episodic layer stores first-person scene memories.
- The semantic layer stores visibility-tagged facts.
- The personality layer stores situation-dependent speech and behavior patterns.
- The authors create a 4,386-question benchmark called KBF-QA over eight novels to evaluate the knowledge boundaries of their method.
- REVERIEMEM shows improved Knowledge Boundary Fidelity by 34.6 percentage points compared to the strongest previous method.
Key Concepts Mentioned
- book-based role-playing agents
- memory architecture
- factual overreach
- stylistic monotony
- knowledge boundaries
- evaluation benchmark
Claims Made by the Document
- REVERIEMEM improves Knowledge Boundary Fidelity by 34.6 percentage points.
- REVERIEMEM achieves a ~79% win rate in BOOKWORLD's five-dimension pairwise narrative protocol.
Evidence Used
- Improved Knowledge Boundary Fidelity by 34.6 percentage points over the strongest prior method.
- ~79% win rate in BOOKWORLD's five-dimension pairwise narrative protocol.
Methods / Approach
UNKNOWN
Limitations Stated by the Authors
UNKNOWN
Limitations Noted by the Agent
Agent interpretation — not from the source.
UNKNOWN
Open Questions
UNKNOWN
Notes for Future Q&A
May include interpretation. Labeled where it does.
UNKNOWN
Raw Abstract / Excerpt
Recent LLM role-playing systems build character agents from novels by extracting characters, scenes, and relations. Yet long-narrative role-playing suffers from two failures: Factual Overreach, where shared retrieval or parametric memory lets a character use facts outside its perspective, and Stylistic Monotony, where profile descriptions flatten a character into a fixed voice. To address these failures, we propose REVERIEMEM, a three-layer memory architecture for book-based character agents. The episodic layer stores first-person scene memories; the semantic layer stores visibility-tagged facts; and the personality layer stores situation-dependent speech and behaviour patterns. For evaluation, we construct KBF-QA, a 4,386-question benchmark over eight novels for testing knowledge boundaries. REVERIEMEM improves Knowledge Boundary Fidelity by 34.6 percentage points over the strongest prior method. On BOOKWORLD's five-dimension pairwise narrative protocol, REVERIEMEM achieves a ~ 79% win rate, suggesting that perspective-bounded memory improves both boundary fidelity and character-grounded narrative generation.
