Posterior Probability Matters: Doubly-Adaptive Calibration for Neural Predictions in Online Advertising

TL;DR

AdaCalib tackles miscalibration in online advertising by introducing a doubly-adaptive field-level calibration framework. It learns a field-specific isotonic calibration function via equi-frequency binning guided by posterior statistics and employs field-adaptive binning to handle varying data frequencies, integrating seamlessly with neural predictors for online serving. The method yields superior field-level calibration (Field-RCE, LogLoss) while preserving ranking (Field-AUC) and demonstrates meaningful online gains in CVR and GMV, validated on both industrial and public datasets. Overall, AdaCalib provides a scalable, field-conditioned calibration solution that improves posterior accuracy without sacrificing retrieval or bidding performance in real-world advertising systems.

Abstract

Predicting user response probabilities is vital for ad ranking and bidding. We hope that predictive models can produce accurate probabilistic predictions that reflect true likelihoods. Calibration techniques aim to post-process model predictions to posterior probabilities. Field-level calibration -- which performs calibration w.r.t. to a specific field value -- is fine-grained and more practical. In this paper we propose a doubly-adaptive approach AdaCalib. It learns an isotonic function family to calibrate model predictions with the guidance of posterior statistics, and field-adaptive mechanisms are designed to ensure that the posterior is appropriate for the field value to be calibrated. Experiments verify that AdaCalib achieves significant improvement on calibration performance. It has been deployed online and beats previous approach.

Figures (4)

• Figure 1: Learned calibration functions.
• Figure 2: Training and serving of AdaCalib.
• Figure 3: Calibration functions of AdaCalib and NeuCalib.
• Figure 4: The number of field values selecting 5 or 20 bins.