Search papers, labs, and topics across Lattice.
Rec-Distill is introduced as an industrial distillation pipeline to transfer performance gains from large-scale recommendation models (up to 24B parameters and 20K sequence length) to efficient serving models. The pipeline uses decoupled training, black-box distillation, a debiasing mechanism, and a hybrid batch-streaming pipeline. Experiments across real-world recommendation and advertising scenarios show that student models recover over 60% of the teacher's gains, translating to measurable business improvements under industrial constraints.
Bridge the gap between offline model scaling and online deployment in recommendation systems: Rec-Distill enables lightweight student models to capture a substantial portion of the performance gains from massive teacher models.
Large recommendation models have demonstrated substantial potential gains under scaling laws, yet these gains are difficult to realize in industrial recommendation systems because real-world deployment requires lightweight models with strict serving efficiency and latency guarantees. This creates a fundamental gap between offline model scaling and online deployment. In this work, we present Rec-Distill, an industrial distillation pipeline that transfers the performance gains of large-scale recommendation modeling to efficient serving models. Rec-Distill combines large-teacher scaling with student-side transfer optimization through decoupled training, black-box distillation, debiasing mechanism, and a hybrid batch-streaming pipeline for dynamic recommendation environments. Across multiple recommendation and advertising scenarios on real-world platforms, our framework scales teacher models up to 24B dense parameters and 20K behavior sequence length, while enabling lightweight students to recover a substantial portion of teacher gains, with distillation transferability exceeding 60% in the best setting. Extensive offline and online experiments further show that these transferred gains consistently translate into measurable business improvements under industrial constraints. These results demonstrate that Rec-Distill provides a practical framework for distilling large-scale recommendation models into deployable, cost-efficient serving systems, while also establishing a reliable path toward scaling recommendation models to even larger regimes in the future.
