Preprint
Aug 2026
SCoRD: Semantic-Assisted Continual Retriever-Reranker Distillation for LLM-Based Recommendation
Experiments on real-world datasets show that SCoRD enables effective and efficient retriever-reranker co-adaptation and introduces a semantic reasoning assistant that distills the LLM's ability to infer underlying user intents into reusable intent-level guidance.
S. Baek, Gyuseok Lee, Seunghan Lee et al.
· 0 citations