Training-free method improves LLM recommendation by profiling both users and items

Original: HARP: Training-Free Dual-Profile Agentic Communication for LLM-Based Recommendation

Source: arxiv.org ↗