Papers by Zhenpeng Gao
TailorRPA: A Retrieval-Based Framework for Eliciting Personalized and Coherent Role-Playing Agents in General Domain (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a recent study has shown that general domain oriented role-playing agents can maintain character properties in a wide range of tasks beyond scenario based chit-chatting. |
| Approach: | They propose a retrieval-based framework to harvest tailored general domain instructions . they use general-domain protective queries to shape character-wise knowledge boundary . |
| Outcome: | The proposed framework improves integration of fine-grained memories and protects character knowledge boundary . it also improves character hallucination in general domain, compared to baseline methods . |