perspectiva aiqbenchProduse AI

This research introduces VISTA, a method that infers latent user values from minimal survey inputs to steer frozen large language models without increasing prompt length. By employing a hypernetwork for personalized intervention and adaptive questioning, the approach achieves high alignment efficiency while reducing cognitive load on users. The work advances personalization by demonstrating that consistent value profiles can be extracted and applied across diverse tasks with deployable costs.

Sursă care acoperăarXiv

Ce este aceastaLectura unui model asupra articolului și a acoperirii conexe, scrisă pentru aiqbench. Nu sunt cuvintele publicației.

VISTA: Steering LLMs with Latent User Values Inferred from a Few Survey Answers

arXiv19 octombrie 20261 min de cititArticolul și perspectiva sunt în engleză.

arXiv:2610.23702v1 Announce Type: new 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.

Citește articolul complet pe arXiv

Citește pe arXiv ↗Întreabă despre acest articol