SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling

TL;DR

SOLARIS is a framework that precomputes likely user-item interactions to enable faster inference in recommendation systems, improving real-time serving without sacrificing model quality.

Contribution

It introduces a speculative offloading method that precomputes embeddings, decoupling expensive inference from latency-critical serving, and demonstrates scalability in a large-scale deployment.

Findings

Achieves 0.67% revenue increase in Meta's advertising system

Enables real-time inference for complex foundation models

Decouples inference from serving latency constraints

Abstract

Recent advances in recommendation scaling laws have led to foundation models of unprecedented complexity. While these models offer superior performance, their computational demands make real-time serving impractical, often forcing practitioners to rely on knowledge distillation-compromising serving quality for efficiency. To address this challenge, we present SOLARIS (Speculative Offloading of Latent-bAsed Representation for Inference Scaling), a novel framework inspired by speculative decoding. SOLARIS proactively precomputes user-item interaction embeddings by predicting which user-item pairs are likely to appear in future requests, and asynchronously generating their foundation model representations ahead of time. This approach decouples the costly foundation model inference from the latency-critical serving path, enabling real-time knowledge transfer from models previously considered too expensive for online use. Deployed across Meta's advertising system serving billions of daily requests, SOLARIS achieves 0.67% revenue-driving top-line metrics gain, demonstrating its effectiveness at scale.