MediumData

Defend your RAG choice and your hallucination eval

1Give yourself 5 minutes

2Answer out loud, not in your head

3Then compare with the answer below

Reference answer

Explain that fine-tuning was cost-prohibitive on the dataset, so we used MiniLM for retrieval speed and a semantic-chunking strategy that dropped hallucination rate by 40%. Present the eval: an LLM-judge against a held-out gold set, paired with human spot-check on a stratified subset, plus retrieval-precision (fraction of retrieved chunks used in the final answer) and factual-accuracy pass rate. Slice by topic to surface where retrieval fails.

Then expect these follow-ups

What about groundedness in borderline cases; when would you switch to FT?

Free to read · better with Enzo

Practice this out loud with Enzo

Enzo runs it as a mock interview, pushes back with follow-ups, and grades you on the rubric.

Continue with Enzo Sign up free

Next question

Bias-variance trade-off
Bagging vs boosting vs Random Forest
Handling imbalanced data
Metric choice for a 0.1%-positive fraud model
Supervised vs unsupervised vs reinforcement learning
Closed-form linear regression vs gradient descent