From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918521329418240 |
|---|---|
| author | Zhang, Rongsheng Hu, Ruofan Chen, Weijie Tang, Jiji Ren, Junnan Wu, Wanying Chen, Xunuoyan Lv, Tangjie Jin, Tao Zhao, Zhou |
| author_facet | Zhang, Rongsheng Hu, Ruofan Chen, Weijie Tang, Jiji Ren, Junnan Wu, Wanying Chen, Xunuoyan Lv, Tangjie Jin, Tao Zhao, Zhou |
| contents | While role-playing agents excel in short-term interactions, long-term conversations overwhelm context windows, motivating external memory frameworks. Current systems typically rely on persona-agnostic summarization, which records facts without persona-specific interpretation, yielding generic responses that compromise persona fidelity. To bridge this gap, we introduce RoleMemo, a dataset featuring four reasoning tasks where the factual fragments must be interpreted through the persona to reach the correct answer. Evaluation on RoleMemo exposes critical limitations of persona-agnostic frameworks. We thus propose DualMem, which decouples memory into two streams: factual cognition and persona-conditioned insight. Trained through Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), our framework with a 4B-parameter model outperforms zero-shot persona-agnostic frameworks powered by DeepSeek-V3.2 for sustained persona fidelity. Our resources are available at https://github.com/role2026/rolememo. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_25693 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents Zhang, Rongsheng Hu, Ruofan Chen, Weijie Tang, Jiji Ren, Junnan Wu, Wanying Chen, Xunuoyan Lv, Tangjie Jin, Tao Zhao, Zhou Computation and Language Databases Multiagent Systems While role-playing agents excel in short-term interactions, long-term conversations overwhelm context windows, motivating external memory frameworks. Current systems typically rely on persona-agnostic summarization, which records facts without persona-specific interpretation, yielding generic responses that compromise persona fidelity. To bridge this gap, we introduce RoleMemo, a dataset featuring four reasoning tasks where the factual fragments must be interpreted through the persona to reach the correct answer. Evaluation on RoleMemo exposes critical limitations of persona-agnostic frameworks. We thus propose DualMem, which decouples memory into two streams: factual cognition and persona-conditioned insight. Trained through Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), our framework with a 4B-parameter model outperforms zero-shot persona-agnostic frameworks powered by DeepSeek-V3.2 for sustained persona fidelity. Our resources are available at https://github.com/role2026/rolememo. |
| title | From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents |
| topic | Computation and Language Databases Multiagent Systems |
| url | https://arxiv.org/abs/2605.25693 |