Research map / Learning theory and algorithms / Prediction and decision losses
Calibration Matters: Tackling Maximization Bias in Large-scale Advertising Recommendation Systems
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Apply VAD to Vanilla predictions | Logistic regression · Synthetic data, α=2% | calibration error: 8.55%$\pm$0.68% -> 0.06%$\pm$0.72% | Helped |
| Apply VAD to Vanilla predictions | Logistic regression · Synthetic data, α=2% | ECE: 0.0656$\pm$0.0019 -> 0.0572$\pm$0.0013 | Helped |
| Apply VAD to Vanilla predictions | DLRM · Criteo Ad Kaggle, α=2% | calibration error: 3.23%$\pm$0.41% -> -0.44%$\pm$0.46% | Helped |
| Apply VAD to Vanilla predictions | DLRM · Criteo Ad Kaggle, α=10% | calibration error: 3.94%$\pm$0.63% -> -0.47%$\pm$0.66% | Helped |
| Combine VAD with Platt Scaling | DLRM · Criteo Ad Kaggle, α=2% | calibration error: 1.60%$\pm$0.13% -> 0.86%$\pm$0.12% | Helped |
| Combine VAD with Platt Scaling | DLRM · Criteo Ad Kaggle, α=10% | ECE: 0.0121$\pm$0.0003 -> 0.0092$\pm$0.0002 | Helped |
| Use identity link instead of sigmoid link (ablation) | VAD · Synthetic data, α=2% | calibration error: 0.06%$\pm$0.72% -> 1.00\%$\pm$0.70\% | Mixed |
| Increase replications from S=2 to S=3 (ablation) | VAD · Synthetic data | ECE: 0.0572$\pm$0.0013 -> 0.0555$\pm$0.0013 | Helped |
| Increase replications from S=2 to S=3 (ablation) | VAD · Criteo Ad Kaggle | calibration error | Mixed |
Closest papers by meaning
- Fine-Grained Dynamic Framework for Bias-Variance Joint Optimization on Data Missing Not at Random (NeurIPS 2024 · similarity 0.86)
- Error-Guided Bellman Calibration for Semi-Offline Value Estimation (NeurIPS 2026 · similarity 0.86)
- Entire-Space Variational Information Exploitation for Post-Click Conversion Rate Prediction (AAAI 2025 · similarity 0.84)
- Addressing Hidden Confounding with Heterogeneous Observational Datasets for Recommendation (NeurIPS 2024 · similarity 0.84)
- TranSUN: A Preemptive Paradigm to Eradicate Retransformation Bias Intrinsically from Regression Models in Recommender Systems (NeurIPS 2025 · similarity 0.83)
- Preference Diffusion for Recommendation (ICLR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.