CAMIE: Co-Engagement-Aware Multimodal Item Embeddings for Snap Dynamic Product Ads Retrieval
2026-08-31 • Information Retrieval
Information Retrieval
AI summaryⓘ
The authors address challenges in item-to-item retrieval within Snap's Dynamic Product Ads by introducing CAMIE, a new system that uses combined image and text data more effectively. CAMIE leverages large language and multimodal models to create shared embeddings for items and fine-tunes them using real user interaction data. Compared to existing models, CAMIE performs better in offline tests and improves click-through and conversion rates when deployed. The system can also handle text-only searches with little loss in quality. Overall, the authors demonstrate that CAMIE can replace their older encoders and improve ad performance in production.
item-to-item retrievalmultimodal embeddinglarge language modelInfoNCE objectiveco-engagementclick-through rateconversion ratedynamic product adsembedding fine-tuningSnap
Authors
Xiaodong Liu, Siman Wang, Congfei Zhang, Hsiang-wei Chao, Xiao Bai, Wen Zhang, Jingxiao Ma, Zhe Liu, Yunzhi Zhou, Yajun Wang, Jinchao Li, Yu Zhang
Abstract
Item-to-item (I2I) retrieval is a core primitive in large-scale recommendation and advertising systems. In production Snap Dynamic Product Ads (DPA), I2I retrieval faces two challenges: separate visual, textual, and multimodal encoders fragment the retrieval stack, and content-only training does not align embeddings with the co-engagement behavior that drives downstream conversions. We present CAMIE, a co-engagement-aware multimodal item embedding framework for Snap DPA retrieval. CAMIE builds on LLM/MLLM backbones, using their native multimodal interfaces to represent item images and metadata in a shared embedding space. It then fine-tunes the backbone on co-engaged item pairs mined from user journeys with a symmetric in-batch InfoNCE objective. Offline, CAMIE outperforms the strongest commercial multimodal embedding model on Recall@10 and serves text-only retrieval from the same checkpoint with minimal quality loss. Online, CAMIE serves as a drop-in replacement for two deployed content-based I2I encoders, delivering +0.390% CTR / +10.832% CVR over the multimodal control, +18.958% CTR / +13.12% CVR over the text control, and +0.211% CTR / +1.911% CVR on overall DPA traffic. CAMIE is deployed in production.