Jul 2026
Candidate-Constrained Retrieval-Augmented Generation for LongEval-RAG: System Design and Empirical Analysis
The result suggests that the main gain does not come from more complex semantic or topic-shift chunking, but from pairing stable rule-based evidence units with sentence-level neural selection before generation, highlighting the need for multi-metric RAG evaluation.
Yingdong Yang, Haijian Wu
· arXiv.org · 0 citations