Training the long-horizon CVR head on fresh short-window clicks and the delay CDF on late arrivals via pCVR-weighted temporal convolution yields monotone CVR predictions at every horizon up to 30 days and a 2.5% expected-revenue lift in production.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TWICE: Two-Clock, Two-Window Learning for Long-Horizon Conversion Prediction in Online Advertising
Training the long-horizon CVR head on fresh short-window clicks and the delay CDF on late arrivals via pCVR-weighted temporal convolution yields monotone CVR predictions at every horizon up to 30 days and a 2.5% expected-revenue lift in production.