Two papers detail RAG systems for LongEval citation integrity task

One paper examines the divergence between traditional NLP evaluation metrics and citation integrity in scientific QA RAG. Another presents a candidate-constrained system that restricts evidence to organizer-provided candidate sets.