arXiv cs.CL· Yinuo Xu, Jadie Adams, Brian Hu, Shivani Kumar, David Jurgens, Arslan Basharat·· 12 小时前AI 评分34
VISTA：从少量问卷答案推断潜在用户价值观来引导 LLM
VISTA: Steering LLMs with Latent User Values Inferred from a Few Survey Answers
AI 导读
VISTA 通过用户（完整或部分）问卷答案推断潜在价值观画像，再经超网络映射为对冻结 LLM 的用户专属引导干预，用于选择答案选项或生成自由文本回复。
正文 · 原文
Abstract:Personalizing a language model requires knowing what a person values, which stays largely consistent across tasks. Yet most methods either learn task-specific preferences or place the user's profile in the prompt, where models struggle to act on it. We introduce VISTA (Value-Informed Steering and Targeted Acquisition), which infers a latent profile from users' (full or partial) survey answers, and maps the profile through a hypernetwork to a user-specific steering intervention in a frozen LLM to select answer choices or generate free-text responses. Given the same evidence, VISTA matches or beats in-context profiles on synthetic preferences (Food2D), the World Values Survey (WVS), and PrimeX, without adding prompt tokens. To reduce the cognitive burden of completing full surveys, VISTA also chooses which questions to ask and when to stop: on WVS it stops after 6% of the questionnaire while retaining 94% of the full-history gain, where standard adaptive testing retains at most 42%. The dominant steering direction is an interpretable value axis on which VISTA places unseen users accurately. Our results point toward pluralistic alignment that adapts to individuals at deployable cost.
| Comments: | 8 pages, 4 figures |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.23702 [cs.CL] |
| (or arXiv:2610.23702v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.23702 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yinuo Xu [view email]
[v1]
Fri, 16 Oct 2026 16:28:57 UTC (451 KB)
来源：arXiv cs.CL · arxiv.org