2026
Align³GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation
AAAI 2026technical
Large Language Models (LLMs) demonstrate significant advantages in leveraging structured world knowledge and multi-step reasoning capabilities. However, fundamental challenges arise when transforming LLMs into real-world recommendation systems due to semantic and behavioral misalignment. To bridge