Practical Multi-Task Learning for Rare Conversions in Ad Tech

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dishi, Yuval, Friedler, Ophir, Karni, Yonatan, Silberstein, Natalia, Stolin, Yulia
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916866907176960
author Dishi, Yuval
Friedler, Ophir
Karni, Yonatan
Silberstein, Natalia
Stolin, Yulia
author_facet Dishi, Yuval
Friedler, Ophir
Karni, Yonatan
Silberstein, Natalia
Stolin, Yulia
contents 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).
format Preprint
id arxiv_https___arxiv_org_abs_2507_20161
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Practical Multi-Task Learning for Rare Conversions in Ad Tech
Dishi, Yuval
Friedler, Ophir
Karni, Yonatan
Silberstein, Natalia
Stolin, Yulia
Information Retrieval
Machine Learning
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).
title Practical Multi-Task Learning for Rare Conversions in Ad Tech
topic Information Retrieval
Machine Learning
url https://arxiv.org/abs/2507.20161