<ul><li>Accurately predicting the conversion rate in online advertising is important for efficiency and user satisfaction.</li><li>A new model training framework has been proposed to predict CVR while preserving user privacy and meeting advertiser requirements.</li><li>The framework uses batch-level aggregated gradients, adapter-based fine-tuning, and de-biasing techniques to minimize privacy risks and reduce communication costs.</li><li>Experimental results show competitive performance with decreased communication overhead, ensuring compliance with privacy requirements in digital advertising.</li></ul>

Privacy Preserving Conversion Modeling in Data Clean Room

Discover more