Research

Paper

TESTING March 03, 2026

PinCLIP: Large-scale Foundational Multimodal Representation at Pinterest

Authors

Josh Beal, Eric Kim, Jinfeng Rao, Rex Wu, Dmitry Kislyuk, Charles Rosenberg

Abstract

While multi-modal Visual Language Models (VLMs) have demonstrated significant success across various domains, the integration of VLMs into recommendation and retrieval systems remains a challenge, due to issues like training objective discrepancies and serving efficiency bottlenecks. This paper introduces PinCLIP, a large-scale visual representation learning approach developed to enhance retrieval and ranking models at Pinterest by leveraging VLMs to learn image-text alignment. We propose a novel hybrid Vision Transformer architecture that utilizes a VLM backbone and a hybrid fusion mechanism to capture multi-modality content representation at varying granularities. Beyond standard image-to-text alignment objectives, we introduce a neighbor alignment objective to model the cross-fusion of multi-modal representations within the Pinterest Pin-Board graph. Offline evaluations show that PinCLIP outperforms state-of-the-art baselines, such as Qwen, by 20% in multi-modal retrieval tasks. Online A/B testing demonstrates significant business impact, including substantial engagement gains across all major surfaces in Pinterest. Notably, PinCLIP significantly addresses the "cold-start" problem, enhancing fresh content distribution with a 15% Repin increase in organic content and 8.7% higher click for new Ads.

Metadata

arXiv ID: 2603.03544
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-03-03
Fetched: 2026-03-05 06:06

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.03544v1</id>\n    <title>PinCLIP: Large-scale Foundational Multimodal Representation at Pinterest</title>\n    <updated>2026-03-03T21:57:16Z</updated>\n    <link href='https://arxiv.org/abs/2603.03544v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.03544v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>While multi-modal Visual Language Models (VLMs) have demonstrated significant success across various domains, the integration of VLMs into recommendation and retrieval systems remains a challenge, due to issues like training objective discrepancies and serving efficiency bottlenecks. This paper introduces PinCLIP, a large-scale visual representation learning approach developed to enhance retrieval and ranking models at Pinterest by leveraging VLMs to learn image-text alignment. We propose a novel hybrid Vision Transformer architecture that utilizes a VLM backbone and a hybrid fusion mechanism to capture multi-modality content representation at varying granularities. Beyond standard image-to-text alignment objectives, we introduce a neighbor alignment objective to model the cross-fusion of multi-modal representations within the Pinterest Pin-Board graph. Offline evaluations show that PinCLIP outperforms state-of-the-art baselines, such as Qwen, by 20% in multi-modal retrieval tasks. Online A/B testing demonstrates significant business impact, including substantial engagement gains across all major surfaces in Pinterest. Notably, PinCLIP significantly addresses the \"cold-start\" problem, enhancing fresh content distribution with a 15% Repin increase in organic content and 8.7% higher click for new Ads.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <published>2026-03-03T21:57:16Z</published>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Josh Beal</name>\n    </author>\n    <author>\n      <name>Eric Kim</name>\n    </author>\n    <author>\n      <name>Jinfeng Rao</name>\n    </author>\n    <author>\n      <name>Rex Wu</name>\n    </author>\n    <author>\n      <name>Dmitry Kislyuk</name>\n    </author>\n    <author>\n      <name>Charles Rosenberg</name>\n    </author>\n  </entry>"
}