arXiv subjects

Zhao Zhu

Publications and source records attributed to Zhao Zhu.

4 recordsLinked to original sources

arXiv2026

Bending the Scaling Law Curve in Large-Scale Recommendation Systems

Learning from user interaction history through sequential models has become a cornerstone of large-scale recommender systems. Recent advances in large language models have revealed promising scaling laws, sparking a surge of research into long-sequence modeling and deeper architectures for recommendation tasks. However, many recent approaches rely heavily on cross-attention mechanisms to address the quadratic computational bottleneck in sequential modeling, which can limit the representational power gained from self-attention. We present ULTRA-HSTU, a novel sequential recommendation model developed through end-to-end model and system co-design. By innovating in the design of input sequences, sparse attention mechanisms, and model topology, ULTRA-HSTU achieves substantial improvements in both model quality and efficiency. Comprehensive benchmarking demonstrates that ULTRA-HSTU achieves remarkable scaling efficiency gains -- over 5x faster training scaling and 21x faster inference scaling compared to conventional models -- while delivering superior recommendation quality. Our solution is fully deployed at scale, serving billions of users daily and driving significant 4% to 8% consumption and engagement improvements in real-world production environments.

cs.IR↗

arXiv2025

Historical Information Accelerates Decentralized Optimization: A Proximal Bundle Method

Historical information, such as past function values or gradients, has significant potential to enhance decentralized optimization methods for two key reasons: first, it provides richer information about the objective function, which also explains its established success in centralized optimization; second, unlike the second-order derivative or its alternatives, historical information has already been computed or communicated and requires no additional cost to acquire. Despite this potential, it remains underexploited. In this work, we employ a proximal bundle framework to incorporate the function values and gradients at historical iterates and adapt the framework to the proximal decentralized gradient descent method, resulting in a Decentralized Proximal Bundle Method (DPBM). To broaden its applicability, we further extend DPBM to the asynchronous and stochastic setting. We theoretically analysed the convergence of the proposed methods. Notably, both the asynchronous DPBM and its stochastic variant can converge with fixed step-sizes that are independent of delays, which is superior to the delay-dependent step-sizes required by most existing asynchronous optimization methods, as it is easier to determine and often leads to faster convergence. Numerical experiments on classification problems demonstrate that by using historical information, our methods yield faster convergence and stronger robustness in the step-sizes.

math.OC↗

arXiv2025

Target-Aware Early Stage Ranking

Early Stage Ranking (ESR) in large-scale recommendation systems is dominated by ''user--item decoupling'' Two Tower architectures, which scale efficiently but cannot capture fine-grained, target-aware user--item interactions directly. We propose Target-Aware Early Stage Ranking (TESR), which augments the Two Tower with a Mixture of Attention (MoA) module trained as a request-level sequence modeling over user history. MoA combines (i) Hard Matching Attention (HMA) to capture explicit categorical-ID level overlap signals between user history and candidate item, (ii) target-aware HSTU attention for implicit affinities conditioned on the candidate, and (iii) target dependent and independent cross-attention for symmetric user-item contextualization. On top of this, a Multi-Logit Parameterized Gating (MLPG) head amplifies these signals at scoring time. To keep latency within ESR budgets, we co-design the architecture with FP8 quantization, custom kernels, and a Torch Inductor compilation path. On a production deployment, TESR delivers consistent offline NE wins and online topline gains, and is, to our knowledge, the first deployment of full target-aware attention sequence modeling in an ESR stage at this scale.

cs.LG↗

arXiv2025

Realizing Scaling Laws in Recommender Systems: A Foundation-Expert Paradigm for Hyperscale Model Deployment

Scaling laws have been established for recommender systems, yet efficiently deploying foundation model (FM) across multiple recommendation surfaces remains a major unsolved challenge. Existing methods for transfer learning face fundamental limitations in this setting: knowledge distillation suffers from transfer fidelity degradation in the large-data regime, and static user or item embeddings lack the expressiveness to capture contextualized user-item interactions. We propose the Foundation-Expert paradigm, where a central FM generates target-aware embeddings which are ingested by lightweight surface-specific expert models. Target-aware embeddings are representations that dynamically capture a user's interest in a specific item conditioned on their full interaction history. Unlike knowledge distillation, which transfers FM knowledge as soft labels, our approach provides these embeddings as input features to each expert model, enabling direct interaction with surface-specific representations. This paradigm achieves transfer ratios of 0.64--1.0 from FM to experts, substantially exceeding existing methods. Fully deployed at Meta serving tens of billions of daily requests since 2025, it delivers 0.050% statistically significant online topline metric improvement and 0.359% cumulative gains across multiple surfaces.

cs.IR↗