FinSAgent: Corpus-Aligned Multi-Agent RAG Framework for SEC Filing QA
FinSAgent is a multi-agent retrieval-augmented generation (RAG) framework designed for evidence-grounded question answering over SEC filings. It addresses the problem of prior-corpus misalignment by conditioning both query generation and reranking on the structure and content of regulatory filings. Experimental results show that FinSAgent improves retrieval coverage and answer correctness compared to strong single-agent and multi-agent baselines, as demonstrated on five financial QA benchmarks and in a randomized online user study.
Why it matters: This work advances the reliability and accuracy of financial question answering systems by aligning retrieval and reasoning with the unique structure and evidentiary standards of SEC filings.
Full story at: arXiv Information Retrieval ↗