EvidenceLens helps auditors fact-check financial AI claims with a visual evidence matrix

Large language models are increasingly used to answer questions over annual reports, earnings decks, and analyst notes, yet their outputs remain difficult to verify in high-stakes financial workflows. A fluent answer can blend directly grounded statements, weak synthesis, and unsupported claims across narrative text, tables, and charts. To address this, researchers from several institutions developed EvidenceLens, a visual analytics prototype that treats financial question answering as a claim-evidence alignment problem. The system decomposes an answer into atomic claims, summarizes support composition and confidence, identifies support gaps, and coordinates claim-level inspection with source passages, table cells, and chart regions.

EvidenceLens's core visual representation is a multimodal claim-evidence matrix that makes coverage, contradiction, and modality imbalance immediately visible. To support reproducibility, the team also specifies a JSON-based artifact schema, a lightweight multimodal alignment pipeline, and a deterministic review-priority ranking that maps backend signals into an auditable visual structure. Through representative report-auditing scenarios, EvidenceLens helps analysts distinguish grounded claims from overconfident synthesis that conventional chat interfaces flatten. The tool is especially relevant for financial auditors who need to verify AI-generated insights before acting on them.