Grokking LLM Evals
Vote

0% completed

​

Red Teaming and the Regression Suite

A red team exercise is a planned, time-limited effort in which people try to make a feature break its policy by searching for inputs that no existing test covers. The people doing it, the red team, act as attackers and as unusual or careless users. Their job is to find failures, not to measure a rate.

A regression suite is the opposite kind of check. It is the set of cases the feature already passes and must keep passing, run unchanged on every change, as Kinds of Evals described

.....

.....

.....

Like the course? Get enrolled and start learning!

Reading Progress

0%


Vote for new content