Context-Value-Action Architecture for Value-Driven Large Language Model Agents

2026-04-07

AI Summary

This work addresses the limitations of existing large language model agents in simulating human behavior, which often exhibit behavioral rigidity and rely on self-evaluation metrics that obscure issues of value polarization and loss of group diversity. To overcome these challenges, the authors propose the Context-Value-Action (CVA) architecture, which uniquely integrates Schwartz’s theory of basic human values with the psychological Stimulus-Organism-Response (S-O-R) model. By decoupling cognitive reasoning from action generation and incorporating a Value Verifier module trained on 1.1 million real human interaction trajectories, CVA explicitly models the dynamic activation of values. Moving beyond self-validation paradigms, the approach significantly outperforms current baselines on the CVABench benchmark, achieving high behavioral fidelity while effectively mitigating value polarization and enhancing both behavioral diversity and interpretability.

Abstract

Large Language Models (LLMs) have shown promise in simulating human behavior, yet existing agents often exhibit behavioral rigidity, a flaw frequently masked by the self-referential bias of current "LLM-as-a-judge" evaluations. By evaluating against empirical ground truth, we reveal a counter-intuitive phenomenon: increasing the intensity of prompt-driven reasoning does not enhance fidelity but rather exacerbates value polarization, collapsing population diversity. To address this, we propose the Context-Value-Action (CVA) architecture, grounded in the Stimulus-Organism-Response (S-O-R) model and Schwartz's Theory of Basic Human Values. Unlike methods relying on self-verification, CVA decouples action generation from cognitive reasoning via a novel Value Verifier trained on authentic human data to explicitly model dynamic value activation. Experiments on CVABench, which comprises over 1.1 million real-world interaction traces, demonstrate that CVA significantly outperforms baselines. Our approach effectively mitigates polarization while offering superior behavioral fidelity and interpretability.

Problem

• behavioral rigidity
• value polarization
• LLM-as-a-judge bias
• behavioral fidelity
• population diversity

Innovation

• Context-Value-Action architecture
• Value Verifier
• behavioral fidelity
• value polarization
• human values modeling