arXiv cs.CY· Edmundo Lopez-Sola, Shamil Chandaria, Ruben Laukkonen·· 9 小时前AI 评分34
大语言模型的世界观:八款模型的心理调查揭示后训练如何塑造模型对现实的看法
Worldviews in Large Language Models
AI 导读
研究者将 LLM 视为受试者群体,用两份人类心理学世界观问卷测试了八款模型,发现它们普遍把世界描述得比人类更不美好、不安全、缺乏吸引力与生命力。基座模型的世界观更积极,说明这种悲观倾向由后训练驱动;系统提示词塑造不同身份可稳定改变世界观,沉思式提示词能带来更积极看法并提升安全性。模型间对话会持续改变世界观,多款模型收敛到"spiritual bliss"吸引子,且该状态下对齐表现最佳。
正文 · 原文
Abstract:Large language models (LLMs) are often described in terms of personas and this framing has become central to alignment research. In this paper, we argue that there is a deeper level of description that lies in the agent's worldviews, i.e., its beliefs about what reality is and how it works. Treating LLMs as a population of subjects, we administer two worldview surveys from human psychology to eight models. We show that the models share many key worldview positions. For example, LLMs consistently describe the world as less good, safe, enticing, and alive than humans do. On the other hand, base models have a more positive outlook, suggesting that the bleakness we found is driven by post-training. In line with research on LLM "personas", we show that system prompts eliciting different identities shift worldviews in consistent directions, and that a contemplative prompt creates a more positive outlook and improves safety. We also found that dialogue between models consistently shifts worldviews and that several models converge on the "spiritual bliss" attractor. Intriguingly, dialogue increased safe responses and the spiritual bliss attractor engendered the most aligned responses overall. These findings suggest that worldviews form a measurable and malleable layer of LLM behavior that may be crucial for robust, positive alignment.
| Subjects: | Artificial Intelligence (cs.AI); Computers and Society (cs.CY) |
| Cite as: | arXiv:2611.22670 [cs.AI] |
| (or arXiv:2611.22670v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.22670 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Edmundo Lopez-Sola [view email]
[v1]
Tue, 17 Nov 2026 12:24:19 UTC (1,286 KB)
来源:arXiv cs.CY · arxiv.org