Show HN: TreasuryBench – an open benchmark for personal-finance AI advice

Category: ai-ml

Tags: benchmark, personal-finance, ai-evaluation

Score: 7.0/10 (Innovation: 6, Technical: 7, Documentation: 8, Utility: 7)

TreasuryBench is an open benchmark for evaluating personal-finance AI assistants across realistic household scenarios, using synthetic personas and fact-checked scoring. It is interesting because it reveals critical gaps in factual accuracy and financial safety for leading products like ChatGPT, alongside a structured methodology and downloadable artifacts.

Target audience: ai researchers, ml engineers, fintech product teams

Repository: https://github.com/Treasury-Technologies-Inc/treasurybench · TypeScript · MIT

View on Hacker News