Asked about Entropy, cross-entropy, KL and perplexity
Where would you meet KL divergence in an LLM pipeline?
An answer that works
As the drift penalty in preference training, keeping the tuned model near the one it started from. As the distillation loss against a teacher's full output distribution. And in monitoring, comparing the distribution of live inputs against the one the model was validated on.
Reading an answer is not being able to give one. The section that teaches this is Entropy, cross-entropy, KL and perplexity, 4 hours of Probability & statistics.
Learn this properlyAlso asked about Probability & statistics
Your new prompt scores 86 percent against 82 percent on a 50-case eval. Ship it?A test for a condition affecting 1 in 1000 people is 99 percent accurate. You test positive. What is the chance you have it?Why is accuracy misleading on imbalanced data?Why report p95 latency rather than the mean?What is maximum likelihood estimation?Why is cross-entropy the loss for classification?How would you put a confidence interval on an eval score?What does it mean for a model to be calibrated?
All 20 questions on Probability & statistics