Unverified
2026
Log-Correlated Extreme-Value Logit Regularizer
Calibrate the maximum attention logit in each head against the log-correlated extreme-value law instead of applying fixed clipping or a fixed max-norm penalty. Penalize only maxima that exceed the predicted log N minus three-quarter log log N baseline by an unusually large order-one fluctuation, allowing ordinary sharp attention while suppressing rare pathological spikes.
Paper: Black Holes and Random Variables
arXiv:2607.02233