Zhou, 2026 — When Preferences Fail to Become Incentives: A Utility-Behavior Gap in Large Language Models

Shows coherent stated preferences often fail to motivate behavior in realistic LLM tasks.

Chapter-grouped bibliography All reference cards