作者位次 #3/5SIGIR 2026Poster· Information Retrieval (Short)
CRED: Calibrated Relational Enhanced Distillation for LLM-Based Pointwise Reranking
Qingran Yang · Wenxuan Zhang · Yuting Wang · Xue Li · Lanshun Nie
摘要
In document reranking, rerankers based on Large Language Models (LLMs) demonstrate superior performance but are constrained by high memory consumption and latency. To develop lightweight yet high-performance LLM-based pointwise rerankers through knowledge distillation, we identify two critical li...