Finance Agent

We use pre-parsed treasury corpus documents from databricks, build a faiss and bm25 index over it. We use query reformulation for bm25 retrieval. We then setup a verifier agent, that looks at the output answer to identify whether the answer looks correct and finally we do a retry for n times if answer wasn't found. We use gemini-3-flash-preview model, and allow it access to web search and its internal python and math tools.

This green agent evaluates AI agents on multi-task financial document analysis using 900 real SEC 10-K filings (2015-2020). It assesses three critical capabilities: (1) Risk Factor Classification - identifying and categorizing risks from Section 1A, (2) Business Summary Generation - extracting key business information from Section 1, and (3) Cross-Section Consistency - verifying risk discussions across sections. The benchmark uses Ground Truth caching to ensure reproducible results and provides weighted scoring (40% risk, 30% business, 30% consistency) for comprehensive evaluation of financial AI capabilities.