arXiv：cs.CL· Maan Qraitem, Kate Saenko, Bryan A. Plummer·· 4 小时前AI 评分36
PersonaWeaver：在程序化角色生成中实现超越传统原型可控多样性
PERSONAWEAVER: Controllable Diversity Beyond Conventional Archetypes in Procedural Character Generation
AI 导读
PersonaWeaver 通过将世界构建与行为规范解耦，并借助人工精选的道德立场与对话反应库来建模角色行为，从而缓解 LLM 生成角色行为趋同的问题。在十种真实与奇幻设定及三个 LLM 上，它产生的道德与交互响应分布比既有方法更广，同时使人际语言、回复长度和情感更具多样性，并减少世界属性的原型化组合。代码已开源。
正文
Abstract:Procedural character generation aims to populate games, simulations, and other virtual worlds with diverse characters. Large language models (LLMs) offer a promising foundation for scaling this task. However, LLM-based procedural character generation remains at an early stage: existing methods either generate characters directly or adapt profiles retrieved from persona banks. As we show, both approaches produce behaviorally homogeneous populations: characters overwhelmingly agree with positive moral norms and respond to questions with helpful, assistant-like reactions. To mitigate this homogenization, we introduce PersonaWeaver, which disentangles world building from behavioral specification and models behavior through setting general, diverse, manually curated banks of moral positions and conversational reactions. This design allows us to test how far LLM(s) can be pushed beyond their default behavioral patterns across settings. Across ten realistic and fantastical settings and three LLM(s), PersonaWeaver produces broader moral and interactional response distributions than prior work. Its guidance also diversifies interpersonal language, response length, and sentiment. It also produces less archetypal combinations of world attributes. Code is available at this https URL.
| Comments: | Accepted at the 1st PANDORA Workshop: Pluralistic AI and NLP |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2609.26629 [cs.CL] |
| (or arXiv:2609.26629v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2609.26629 arXiv-issued DOI via DataCite |
Submission history
From: Maan Qraitem [view email]
[v1]
Tue, 22 Sep 2026 16:04:14 UTC (18,422 KB)
[v2]
Tue, 6 Oct 2026 21:20:27 UTC (18,422 KB)
来源：arXiv：cs.CL · arxiv.org