PEI at CheckThat! 2026: Investigating Query Normalization and Retrieve-and-Rerank Strategies for Linking Scientific Web Claims to Publications
This paper proposes a retrieve-and-rerank baseline consisting of the Harrier-27b dense retriever and the Jina cross-encoder reranker, and explores several extensions to improve the baseline, including multilingual query normalization through translation and noise removal, dense retriever and reranker fine-tuning, and l...