AIClassification and calibration in cybersecurity evaluations
A user reports better results in their cybersecurity evaluation but says they don't yet have a full comparison, highlighting calibration as a challenge.
TLDR
A user shared an additional comparison using a separate set of 28 questions. In a follow-up reply, they reported better results in their cybersecurity evaluation, while noting they did not yet have a full comparison. They called getting classification right “unbelievably useful,” but said calibration is hard.
Combined views
393
1 Source, first seen 1d ago
Classification and calibration in cybersecurity evaluations
A user reports better results in their cybersecurity evaluation but says they don't yet have a full comparison, highlighting calibration as a challenge.
TLDR
A user shared an additional comparison using a separate set of 28 questions. In a follow-up reply, they reported better results in their cybersecurity evaluation, while noting they did not yet have a full comparison. They called getting classification right “unbelievably useful,” but said calibration is hard.