Saved in:
Bibliographic Details
Main Authors: Liu, Zikun, Luo, Liang, Li, Qianru, Zhang, Zhengyu, Ling, Wei, Shen, Jingyi, Chen, Zeliang, Huang, Yaning, Huang, Jingxian, Aboelela, Abdallah, Sun, Chonglin, Gu, Feifan, Wu, Fenggang, Qu, Hang, Li, Huayu, Pan, Jill, Pei, Kaidi, Chen, Laming, Jin, Longhao, Huang, Qin, Tang, Tongyi, Puvvada, Varna, Chen, Wenlin, Wei, Xiaohan, Cao, Xu, Yao, Yantao, Jin, Yuan, Pu, Yunchen, Chen, Yuxin, Shen, Zijian, Zhang, Zhengkai, Liang, Dong, Wen, Ellie
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.12110
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913029013110784
author Liu, Zikun
Luo, Liang
Li, Qianru
Zhang, Zhengyu
Ling, Wei
Shen, Jingyi
Chen, Zeliang
Huang, Yaning
Huang, Jingxian
Aboelela, Abdallah
Sun, Chonglin
Gu, Feifan
Wu, Fenggang
Qu, Hang
Li, Huayu
Pan, Jill
Pei, Kaidi
Chen, Laming
Jin, Longhao
Huang, Qin
Tang, Tongyi
Puvvada, Varna
Chen, Wenlin
Wei, Xiaohan
Cao, Xu
Yao, Yantao
Jin, Yuan
Pu, Yunchen
Chen, Yuxin
Shen, Zijian
Zhang, Zhengkai
Liang, Dong
Wen, Ellie
author_facet Liu, Zikun
Luo, Liang
Li, Qianru
Zhang, Zhengyu
Ling, Wei
Shen, Jingyi
Chen, Zeliang
Huang, Yaning
Huang, Jingxian
Aboelela, Abdallah
Sun, Chonglin
Gu, Feifan
Wu, Fenggang
Qu, Hang
Li, Huayu
Pan, Jill
Pei, Kaidi
Chen, Laming
Jin, Longhao
Huang, Qin
Tang, Tongyi
Puvvada, Varna
Chen, Wenlin
Wei, Xiaohan
Cao, Xu
Yao, Yantao
Jin, Yuan
Pu, Yunchen
Chen, Yuxin
Shen, Zijian
Zhang, Zhengkai
Liang, Dong
Wen, Ellie
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2604_12110
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling
Liu, Zikun
Luo, Liang
Li, Qianru
Zhang, Zhengyu
Ling, Wei
Shen, Jingyi
Chen, Zeliang
Huang, Yaning
Huang, Jingxian
Aboelela, Abdallah
Sun, Chonglin
Gu, Feifan
Wu, Fenggang
Qu, Hang
Li, Huayu
Pan, Jill
Pei, Kaidi
Chen, Laming
Jin, Longhao
Huang, Qin
Tang, Tongyi
Puvvada, Varna
Chen, Wenlin
Wei, Xiaohan
Cao, Xu
Yao, Yantao
Jin, Yuan
Pu, Yunchen
Chen, Yuxin
Shen, Zijian
Zhang, Zhengkai
Liang, Dong
Wen, Ellie
Machine Learning
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.
title SOLARIS: Speculative Offloading of Latent-bAsed Representation for Inference Scaling
topic Machine Learning
url https://arxiv.org/abs/2604.12110