From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMs

2026-09-23

AI Summary

Existing personalized alignment methods for large language models predominantly rely on static user profiles, overlooking the context-dependent dynamics of human values. To address this limitation, this work proposes BaCVA, a field-theory-inspired Bayesian Contextual Value Alignment framework. BaCVA formulates individual values as priors and contextual preferences as posteriors, introducing a novel Bayesian approach to estimate context-aware value salience. Furthermore, a dual-perspective module dynamically infers posterior preferences during inference, enabling adaptive integration. Experimental results demonstrate that BaCVA significantly outperforms strong baselines across multiple benchmarks, achieving precise, efficient, and context-adaptive personalized value alignment with superior data efficiency.

Abstract

Personalized value alignment has become increasingly important as large language models (LLMs) are expected to accommodate diverse user preferences. However, existing methods typically align model outputs with a static value profile across prompts, overlooking that the salience of value dimensions varies substantially across contexts. Inspired by Lewin's Field Theory, which views human behavior as jointly shaped by personal dispositions and situational constraints, we model personal values as priors and context-dependent preferences as posteriors. We propose BaCVA, an inference-time Bayesian Context-aware personalized Value Alignment method that approximates posterior personalized preferences by integrating static personal values with scenario-specific value salience. BaCVA first estimates contextual value salience from generally normative responses, and then employs a dual-view personalization module to infer posterior preferences from complementary personal-value and scenario-driven perspectives. This Bayesian formulation enables more accurate and adaptive personalized value alignment while improving data efficiency via prior values. Extensive experiments on benchmarks demonstrate its superiority over strong baselines.

Problem

• Personalized Value Alignment
• Large Language Models
• Contextualized Personalization
• Static Value Profile

Innovation

• Bayesian Value Alignment
• Context-aware Personalization
• Large Language Models
• Inference-time Adaptation
• Dual-view Personalization