Practical Multi-Task Learning for Rare Conversions in Ad Tech

Yuval Dishi, Ophir Friedler, Yonatan Karni, Natalia Silberstein, Yulia Stolin
7/27/2025
cs.IRcs.LG

No confident code implementation found for this paper.

Abstract

We present a Multi-Task Learning (MTL) approach for improving predictions for rare (e.g., <1%) conversion events in online advertising. The conversions are classified into "rare" or "frequent" types based on historical statistics. The model learns shared representations across all signals while specializing through separate task towers for each type. The approach was tested and fully deployed to production, demonstrating consistent improvements in both offline (0.69% AUC lift) and online KPI performance metric (2% Cost per Action reduction).

View on arXivView PDF

Code Implementations

No confident code match yet

We couldn't find an author-owned or strongly-evidenced community implementation for this paper. Any repos shown below are weak matches — verify before relying on them.

No code implementations found yet.

Know of an implementation? Let us know in the comments below!

Similar Papers

No One Left Behind: How to Exploit the Incomplete and Skewed Multi-Label Data for Conversion Rate Prediction

86% match2025

Improving conversion rate prediction via self-supervised pre-training in online advertising

81% match2024

The Evolution of Embedding Table Optimization and Multi-Epoch Training in Pinterest Ads Conversion

80% match2025

Practical Lessons on Optimizing Sponsored Products in eCommerce

79% match2023

Bias Correction in Machine Learning-based Classification of Rare Events

79% match2024

Trajectory-wise Iterative Reinforcement Learning Framework for Auto-bidding

77% match2024

Cite this paper

@article{dishi2025practical,
  title  = {Practical Multi-Task Learning for Rare Conversions in Ad Tech},
  author = {Yuval Dishi and Ophir Friedler and Yonatan Karni and Natalia Silberstein and Yulia Stolin},
  year   = {2025},
  eprint = {2507.20161},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2507.20161v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2507.20161.svg)](https://codeofpaper.com/paper/2507.20161)

Discussion