Update GPT-6 Luna benchmark results
#36
by jfbrown95 - opened
Publish GPT-6 Luna Decisions results for all public and private bug, security and semantic lint benchmarks. Use native two-option questions with the configured score cutoffs. Include public probabilities, aggregate private scores, inference costs and matching graphs.
Validation: all 9,506 outcomes verified, 11 tests passed and tables and graphs regenerate correctly.
jfbrown95 changed pull request status to open
jfbrown95 changed pull request title from Add Decisions choice benchmark to Update GPT-6 Luna benchmark results
jfbrown95 changed pull request status to merged