Item analysis

Question statistics

Attempts · 6Questions with data · 5

Question	n	Difficulty p	Discrimination r_pb	Verdict
How does DDIM achieve faster generation compared to DDPM?	6	0.33	0.96	OK
What problem does the reparametrization trick solve in training a VAE?	6	0.50	0.44	OK
Why does a VAE need a KL divergence term in its loss function, even though it already includes a reconstruction error?	6	0.83	0.18	OK
Why is Stable Diffusion more computationally efficient than pixel-space diffusion models?	6	0.33	0.96	OK
Why is the VAE loss derived from the ELBO considered principled rather than ad‑hoc?	6	0.50	0.71	OK

Rejection rule (from 5 attempts): a question is flagged when its difficulty p falls outside 0.30–0.85 (share answering correctly) or its point-biserial discrimination drops below 0.15 (correlation with the rest of the attempt score). Retiring and regenerating flagged questions runs offline: uv run python -m app.quality … --regenerate.