The number
The bold number in a results table is the field's account of what got better, and nothing in the table says what it was computed against. The bench is built here.
Open session → Lecture 01 · The bench, live →The twelve-week teaching schedule: what happens in each session, what to prepare, and what leaves the room.
SLOP8412 / Teaching Session / Reading
Sessions
Dive in FID to study with concepts, math codes and applications.
Arc A / Weeks 1–4
Compute it
The Fréchet distance
The closed form for two Gaussians, derived rather than quoted, and the assumption it smuggles in
Open session → Lecture 02 · Two Gaussians, One Number →The instrument
Inception-v3 was trained to name a thousand ImageNet classes. The score measures distance in the space it built to do that.
Open session →Getting it right
The same samples can receive different scores. Separate preprocessing from numerical arithmetic, then pin the protocol you can defend.
Open session →Arc B / Weeks 5–8
Break it
The bias
Finite-sample FID is biased, the bias depends on the generator, and fixing N does not repair the comparison
Open session → Lecture 05 · Question the estimate →Blind spots
A generator that memorised the reference set scores perfectly. What else the score cannot see, demonstrated on the bench.
Open session → Codewalk: reproduce the number · 9 September 2026 →Dropping the Gaussian
An unbiased estimator need not detect every difference. Compare KID, density and coverage against the bench's known Gaussian-distance baseline.
Open session →Video
FVD keeps the Gaussian distance but changes the features. Test what a temporal instrument preserves before transferring claims from images.
Open session → Lecture 08 · Change the instrument →Arc C / Weeks 9–12
Live with it
Human evaluation
The judgement the score stands in for. How a rating study is designed, what it costs, and why it is so rarely run.
Open session →Correlation
Put the ranking a score produces beside the ranking people produce, and ask whether they agree. That question decides whether the previous nine weeks were worth doing.
Open session → Codewalk: break the number · 14 October 2026 →Why it persists
Nine years of results tables carry an FID column, and that continuity is the strongest argument anyone has for keeping it. What the number costs to leave out, stated honestly.
Open session → Lecture 11 · Defend the report →What you would report instead
Defend your proposal by the standard built across the semester. Bring the final bench comparison and the conditions under which you would still report FID.
Open session → Course closing · The number is not the conclusion →