Can gradient descent learn around padded zeros?
A five-seed DistilHuBERT experiment asks whether a model can learn around BatchNorm statistics contaminated by padding.
Article · 2026
Garden theme 01
What changes when evidence meets criteria, priors, persuasion, or a rating scale?
A five-seed DistilHuBERT experiment asks whether a model can learn around BatchNorm statistics contaminated by padding.
Article · 2026
A psychology-inspired experiment inside Gemma: strong category responses, a few useful surprises, and a hypothesis that did not survive its controls.
Article · 2026
Why counts and population rates can produce different rankings, and what neither measure can establish alone.
Note · 2026
A Bayesian view of polling and the uncertainty around a US presidential election.
Article · 2025
Mixed-effects linear models and generalized linear models, continued from the Bayesian modeling series.
Article · 2024
An introduction to Bayesian modeling for psychologists, with an emphasis on updating statistical skills.
Article · 2024
When different interpretations of probability disagree about what p-values mean.
Article · 2023