FinGround: Detecting and Grounding Financial Hallucinations via Atomic Claim Verification

TL;DR

FinGround is a three-stage pipeline that verifies and grounds financial claims in documents, significantly reducing hallucinations in financial AI systems by leveraging atomic claim verification and structured retrieval.

Contribution

The paper introduces FinGround, a novel verify-then-ground pipeline with a new evaluation methodology, achieving substantial hallucination reduction in financial document QA.

Findings

FinGround reduces hallucinations by 68% over the strongest baseline.

The full pipeline achieves a 78% reduction relative to GPT-4o.

An 8B distilled detector maintains 91.4% F1 at 18x lower latency.

Abstract

Financial AI systems must produce answers grounded in specific regulatory filings, yet current LLMs fabricate metrics, invent citations, and miscalculate derived quantities. These errors carry direct regulatory consequences as the EU AI Act's high-risk enforcement deadline approaches (August 2026). Existing hallucination detectors treat all claims uniformly, missing 43% of computational errors that require arithmetic re-verification against structured tables. We present FinGround, a three-stage verify-then-ground pipeline for financial document QA. Stage 1 performs finance-aware hybrid retrieval over text and tables. Stage 2 decomposes answers into atomic claims classified by a six-type financial taxonomy and verified with type-routed strategies including formula reconstruction. Stage 3 rewrites unsupported claims with paragraph- and table-cell-level citations. To cleanly isolate verification value from retrieval quality, we propose retrieval-equalized evaluation as standard methodology for RAG verification research: when all systems receive identical retrieval, FinGround still reduces hallucination rates by 68% over the strongest baseline ($p < 0.01$). The full pipeline achieves a 78% reduction relative to GPT-4o. An 8B distilled detector retains 91.4% F1 at 18x lower per-claim latency, enabling $0.003/query deployment, supported by qualitative signals from a four-week analyst pilot.