Benchmark
ConvFinQA
Test numerical reasoning across conversations about financial reports.
Published leaderboard
Explore scores and evaluation settings on the publisher’s leaderboard.
Open leaderboardCoverage
3,892 conversations and 14,115 questions; the official split has 434 test conversations.
Methodology
Questions decompose and combine FinQA reasoning chains. Models generate executable reasoning programs for successive turns while using report evidence and conversation history.