Research

Paper

TESTING March 04, 2026

Constraint-Aware Generative Re-ranking for Multi-Objective Optimization in Advertising Feeds

Authors

Chenfei Li, Hantao Zhao, Weixi Yao, Ruiming Huang, Rongrong Lu, Geng Tian, Dongying Kong

Abstract

Optimizing reranking in advertising feeds is a constrained combinatorial problem, requiring simultaneous maximization of platform revenue and preservation of user experience. Recent generative ranking methods enable listwise optimization via autoregressive decoding, but their deployment is hindered by high inference latency and limited constraint handling. We propose a constraint-aware generative reranking framework that transforms constrained optimization into bounded neural decoding. Unlike prior approaches that separate generator and evaluator models, our framework unifies sequence generation and reward estimation into a single network. We further introduce constraint-aware reward pruning, integrating constraint satisfaction directly into decoding to efficiently generate optimal sequences. Experiments on large-scale industrial feeds and online A/B tests show that our method improves revenue and user engagement while meeting strict latency requirements, providing an efficient neural solution for constrained listwise optimization.

Metadata

arXiv ID: 2603.04227
Provider: ARXIV
Primary Category: cs.IR
Published: 2026-03-04
Fetched: 2026-03-05 06:06

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.04227v1</id>\n    <title>Constraint-Aware Generative Re-ranking for Multi-Objective Optimization in Advertising Feeds</title>\n    <updated>2026-03-04T16:09:36Z</updated>\n    <link href='https://arxiv.org/abs/2603.04227v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.04227v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Optimizing reranking in advertising feeds is a constrained combinatorial problem, requiring simultaneous maximization of platform revenue and preservation of user experience. Recent generative ranking methods enable listwise optimization via autoregressive decoding, but their deployment is hindered by high inference latency and limited constraint handling.\n  We propose a constraint-aware generative reranking framework that transforms constrained optimization into bounded neural decoding. Unlike prior approaches that separate generator and evaluator models, our framework unifies sequence generation and reward estimation into a single network.\n  We further introduce constraint-aware reward pruning, integrating constraint satisfaction directly into decoding to efficiently generate optimal sequences. Experiments on large-scale industrial feeds and online A/B tests show that our method improves revenue and user engagement while meeting strict latency requirements, providing an efficient neural solution for constrained listwise optimization.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.IR'/>\n    <published>2026-03-04T16:09:36Z</published>\n    <arxiv:comment>14 pages, 2 figures, 3 tables</arxiv:comment>\n    <arxiv:primary_category term='cs.IR'/>\n    <author>\n      <name>Chenfei Li</name>\n    </author>\n    <author>\n      <name>Hantao Zhao</name>\n    </author>\n    <author>\n      <name>Weixi Yao</name>\n    </author>\n    <author>\n      <name>Ruiming Huang</name>\n    </author>\n    <author>\n      <name>Rongrong Lu</name>\n    </author>\n    <author>\n      <name>Geng Tian</name>\n    </author>\n    <author>\n      <name>Dongying Kong</name>\n    </author>\n  </entry>"
}