Metadata Matters: A Hybrid Retrieval Framework for Structured Financial Document Analysis
Large Language Models are deployed in financial applications such as research synthesis and risk analysis, yet their effectiveness is constrained by the limitations of conventional retrieval methods. Existing approaches rely primarily on semantic similarity or token-level matching, which fails in structured domains lik...