Search Relevance Prediction

1 post · Jun 2026

What this trail covers

DoorDash implemented a new relevance prediction system for search ads, leveraging small language models (SLMs) to improve the accuracy of ad matching to user search intent. This system uses an LLM for offline label generation and an SLM for low-latency online prediction, employing a bi-encoder architecture with BERT-based encoders and 64-dimensional embeddings. The system now uses a DistilBERT bi-encoder with shared encoder weights, CLS pooling, 64-dimensional embeddings, and an online bilinear layer for low-latency relevance scoring, trained using CORAL for ordinal regression.

All DoorDash trails

Jun 2026Using small language models to serve more relevant DoorDash search ads - DoorDashThis post details the evolution of DoorDash's search relevance prediction system by introducing the use of small language models (SLMs) for low-latency online prediction. It describes the teacher-student paradigm where a fine-tuned LLM generates offline relevance labels (0-1-2 scale) and an SLM learns from these labels. The production model utilizes a DistilBERT bi-encoder with shared weights, CLS pooling, and 64-dimensional embeddings, with an online bilinear layer for scoring. The training process was optimized using the CORAL framework for ordinal regression, achieving 97% validation accuracy. The system is integrated into the serving pipeline to filter irrelevant ads and improve user experience.
Advertisement