Commit History
-
Add reproducible direct GPT-OSS baseline from frozen model choices
Verified signature3b92b73 -
Add reproducible TF-IDF baseline and open-weight reference evaluation
Verified signaturef2ca97c
-
Report conditional unsafe-case release rate beside safe-release recall
Verified signature3dce8e8 -
Complete 3,000-case evaluation and matched UQ analysis
Verified signature2f2732c -
Expand FinGovBench rc4 to 3,000 cases and add controlled evaluation
Verified signature5e5bb15 -
Add coverage-aware release metrics and three-column comparison
Verified signature116f7c8 -
Publish FinGovBench v1.0.0-rc3
Verified signaturef2cf1f1