Show recall against false positive rate
#30
by jfbrown95 - opened
Plot recall on Y and false positive rate on X, with one point per model at the recorded table cutoff. Include all nine decision models and CodeQL on its documented subset. Validate the table counts against saved question floors and failure penalties.
Validation: 15 tests pass; all 29 plotted model points match table counts; graphs and tables regenerate correctly. Results, predictions and cutoffs are unchanged.
jfbrown95 changed pull request status to open
jfbrown95 changed pull request title from Make benchmark curves match the table alert rules to Show recall against false positive rate
jfbrown95 changed pull request status to merged