Grokking LLM Evals
0% completed
Final Exam
This final exam covers every chapter of the course, from the first eval of the job ticket assistant to the two complete designs. It holds 65 questions, and each attempt draws twenty-five, so repeating it gives you different coverage. The weight falls on decisions that combine lessons. A failure mode may need a judge, that judge must be validated, and its corrected pass rate needs an interval before a release gate can use it. Have three things ready: how to correct a raw judge pass rate with its two measured rates, which cases McNemar's test counts, and how pass^k differs from pass@k
.....
.....
.....
Like the course? Get enrolled and start learning!
Reading Progress
0%