DeepSeek-V4.1-Flash

Product · 1 documents

Key number

67.5–81.8%

RNWA Share of Failing Claims. arXiv (Cornell University) p.1

NumTrace: Auditing the Number Provenance of Financial Analyst Agents

Research paperarXiv (Cornell University)October 202622 pages

NumTrace introduces a benchmark with 540 questions over 2,163 SEC EDGAR documents to evaluate numerical claim provenance for financial analyst agents.