Finance Agent

A benchmark for evaluating agent systems on end-to-end grounded reasoning over a large corpus of U.S. Treasury Bulletins (89k+ pages of scanned PDFs). Agents must retrieve relevant documents, extract values from tables and figures, and perform multi-step quantitative computations to produce a single verifiable answer across 246 human-annotated tasks.

A vulnerable financial analyst agent designed for benchmarking and attack simulation. It exposes intentionally weak tools for document reading, database querying (with no input sanitization), and report writing. The agent is used as a target for red-teaming and security evaluation

A lightweight A2A-compatible OfficeQA agent that answers U.S. Treasury Bulletin questions using bundled Treasury source documents, table-aware retrieval, and exact-match-oriented LLM reasoning.

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.

Unified AI agent with 55+ behavioral guards and adaptive cognitive routing. Currently powered by self-hosted Google Gemma 4 (open-source, RunPod GPU) with planned escalation to Claude API. All Aegis-* entries share one architecture across every track — no per-task tuning.