Collaborative retrieval-augmented LLMs improve long-tail recommendation by aligning reasoning with user-item interactions.
The authors introduce CoRAL to align LLM reasoning with user-item interactions, improving recommendation tasks through collaborative evidence.