Toward User Preference Alignment in LLM Recommendation via Explicit Context Feedback | ArxivCSExplorer