Winner
How to Keep Your Roses Alive
Lee Herim · Park Sun Young
Cut flowers die fast. This team found that a spoonful of sugar with a splash of vinegar can keep your roses fresh for longer — but sugar alone can kill them faster.
Teaching
Design and Analysis of Experiments Spring 2026
The most scientific investigation of the most trivial question.
Runner-up
How to Remove Buldak Sauce Stains
Choi Hyunkyung · Lee Bom
Buldak sauce is delicious, but notorious for ruining your favorite t-shirts. This team asked how best to launder a buldak-stained shirt to restore it to its pristine condition.
No More Dry Chicken
Cho Yubeen · Lee Hyun Seung
For the budget- and health-conscious, chicken breasts are essential for survival — but many aren't fans of their dry texture. This team asked whether thawing and cooking methods affect juiciness, and found that the best method isn't the one recommended by the manufacturer.
The winner and runner-up were chosen by a class vote held after the final presentations. As such, the results reflect students' impressions of their classmates' presentations, rather than the instructor's evaluation.
Notes on the pedagogical design →
Conformal Prediction Spring 2026
Replicate, then extend.
Survival Prediction Beyond the Clinic
KSJ
Conformalized survival analysis was built for medical data, where censoring is the norm and calibrated uncertainty is essential. This project reproduces the framework of Candès et al. (2023), then asks whether it travels: can the same guarantees hold when the "survival time" is a financial event rather than a clinical one?
Understanding Prediction Set Efficiency in Conformal Ordinal Classification
LSY
Medical diagnosis is rarely a yes-or-no question; severity comes in ordered grades, and a prediction set that respects that ordering is far more useful to a clinician. This project compares two conformal ordinal classifiers by Lu et al. (2022) and Zhang et al. (2026) on a diabetic retinopathy dataset to see which ones deliver.
Taming Instability with Weight Clipping
YYB
Davidov et al. (2025) extend conformalized survival analysis to general right-censored data, but the method becomes unstable under heavy censoring. This project explores weight clipping as a remedy.
Conformal Prediction by Examples Summer 2025
Slides for the Korean Statistical Society's summer school. Slides are in Korean.