# Compute Only Once: UG-Separation for Efficient Large Recommendation Models

arXiv: https://arxiv.org/abs/2602.10455 · HTML: https://mlsanity.com/paper/2602.10455/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 18 of its 31 references that are in the map.
- First posted on arXiv 11 Feb 2026.

## Builds on (most cited first, 18 of 18)

- [MLP-Mixer: An all-MLP Architecture for Vision](https://mlsanity.com/paper/2105.01601/) (2021)
- [Deep & Cross Network for Ad Click Predictions](https://mlsanity.com/paper/1708.05123/) (2017)
- [Deep Interest Evolution Network for Click-Through Rate Prediction](https://mlsanity.com/paper/1809.03672/) (2018)
- [OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignment](https://mlsanity.com/paper/2502.18965/) (2025)
- [Multi-Interest Network with Dynamic Routing for Recommendation at Tmall](https://mlsanity.com/paper/1904.08030/) (2019)
- [Sequential Recommendation with Graph Neural Networks](https://mlsanity.com/paper/2106.14226/) (2021)
- [Wukong: Towards a Scaling Law for Large-Scale Recommendation](https://mlsanity.com/paper/2403.02545/) (2024)
- [PEPNet: Parameter and Embedding Personalized Network for Infusing with Personalized Prior Information](https://mlsanity.com/paper/2302.01115/) (2023)
- [LONGER: Scaling Up Long Sequence Modeling in Industrial Recommenders](https://mlsanity.com/paper/2505.04421/) (2025)
- [RankMixer: Scaling Up Ranking Models in Industrial Recommenders](https://mlsanity.com/paper/2507.15551/) (2025)
- [An Attentive Inductive Bias for Sequential Recommendation beyond the Self-Attention](https://mlsanity.com/paper/2312.10325/) (2023)
- [Zero-shot Quantization: A Comprehensive Survey](https://mlsanity.com/paper/2505.09188/) (2025)
- [Towards the Better Ranking Consistency: A Multi-task Learning Framework for Early Stage Ads Ranking](https://mlsanity.com/paper/2307.11096/) (2023)
- [UnifiedSSR: A Unified Framework of Sequential Search and Recommendation](https://mlsanity.com/paper/2310.13921/) (2023)
- [Squeeze and Excitation: A Weighted Graph Contrastive Learning for Collaborative Filtering](https://mlsanity.com/paper/2504.04443/) (2025)
- [MTmixAtt: Integrating Mixture-of-Experts with Multi-Mix Attention for Large-Scale Recommendation](https://mlsanity.com/paper/2510.15286/) (2025)
- [Pyramid Mixer: Multi-dimensional Multi-period Interest Modeling for Sequential Recommendation](https://mlsanity.com/paper/2506.16942/) (2025)
- [A Non-Uniform Quantization Framework for Time-Encoding Machines](https://mlsanity.com/paper/2511.02728/) (2025)

## Built on by (most cited first, 1 of 1)

- [ROCS: Request-Oriented Compute Sharing for Efficient Large-Scale Recommendation](https://mlsanity.com/paper/2607.27744/) (2026)

## Abstract (arXiv)

Driven by scaling laws, recommender systems increasingly rely on larger-scale models to capture complex feature interactions and user behaviors, but this trend also leads to prohibitive training and inference costs. While long-sequence models can reuse user-side computation through KV Caching, such reuse is difficult in TokenMixer-based dense feature interaction architectures, where user and group features are deeply entangled and mixed-up across layers. In this work, we present User-Group Separation (UG-Sep), an industrial large-scale framework that enables user-side computation reusable in TokenMixer-based dense interaction models for the first time. UG-Sep explicitly disentangles user-side and item-side information flows within token-mixing layers, ensuring that a subset of tokens preserves purely user-side representations across layers. This design allows the corresponding per-token computations to be reused across multiple samples, significantly reducing redundant inference cost. To compensate for the potential expressive capacity loss induced by masking, we further propose an Information Compensation strategy that adaptively reconstructs suppressed user-item interactions. Moreover, as UG-Sep substantially reduces user-side FLOPs and exposes memory-bound components, we incorporate W8A16 (8-bit weight, 16-bit activation) weight-only quantization to alleviate memory bandwidth bottlenecks and achieve additional acceleration. We conduct extensive offline evaluations and large-scale online A/B experiments at ByteDance to validate the effectiveness of UG-Sep. Results show that UG-Sep reduces inference latency by up to 20% without causing adverse changes to online user experience and commercial metrics on multiple influential business scenarios compared to TokenMixer at ByteDance, including Douyin Feed Recommendation, Hongguo Feed Recommendation, Chuanshanjia Ads, and Qianchuan Ads.