Zhe Zhao

Zhe Zhao

Auto-generated profile, based on publicly available information

Zhe Zhao

Google DeepMind, Mountain View, CA, USA

Research Focus & Metrics

Academic Career Length

16Years (2011-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 October 2023
Verified
Conference Paper
Open access
QLQingyun Liu
Qingyun Liu
ZZZhe Zhao
Zhe Zhao
LLLiang Liu
Liang Liu
ZZZhen Zhang
Zhen Zhang
JSJunjie Shan
Junjie Shan
ECEd H. Chi
Ed H. Chi
In recent years, social media users spend significant amount of time on Short-Form Video (SFV) platforms. Its success in creating an immersive viewership experience is not only from the content, but also due to its unique UI innovation: instead of providing choices for users to click, SFV platforms actively recommend content to users to watch one at a time. In this paper, we highlight unique challenges rooted from such UI changes for SFV recommendation system design. Firstly, there is yet much unexplored for sources of system biases under the new UI, as there are no clicks nor the common click-based position biases. Additionally, when training multiple types of user activities, positive labels for activities like sharing and commenting can be much sparser and more skewed than traditional click-based recommendation systems, as the latter can filter non-click impressions when generating "post-click" activities. To tackle these challenges, we introduce a unified multi-task ranking framework which puts two novel components all together into an overall system for SFV recommendation. First, we identify that there are position biases of SFVs in the recommendation sequence, namely "watch trail biases", and introduce biases correction using trail-related information. Second, to get the most benefits from multi-task learning, especially co-training tasks with extremely skewed and sparse labels, we adapt a disentangle regularization to mitigate task conflicts, introduce loss upweighting for sparse task co-training and adopt a meta-learning algorithm for efficient weight selection. We demonstrate the effectiveness and efficiency of the framework on one of today's largest SFV platforms. Our framework has been deployed to the production system for more than 6 months.

8 October 2024
Verified
Conference Paper
Open access
NKNikhil Khani
Nikhil Khani
LWLi Wei
Li Wei
ANAniruddh Nath
Aniruddh Nath
SAShawn Andrews
Shawn Andrews
SYShuo Yang
Shuo Yang
ECEd Chi
Ed Chi
Knowledge Distillation (KD) is a powerful approach for compressing a large model into a smaller, more efficient model, particularly beneficial for latency-sensitive applications like recommender systems. However, current KD research predominantly focuses on Computer Vision (CV) and NLP tasks, overlooking unique data characteristics and challenges inherent to recommender systems. This paper addresses these overlooked challenges, specifically: (1) mitigating data distribution shifts between teacher and student models, (2) efficiently identifying optimal teacher configurations within time and budgetary constraints, and (3) enabling computationally efficient and rapid sharing of teacher labels to support multiple students. We present a robust KD system developed and rigorously evaluated on multiple large-scale personalized video recommendation systems within Google. Our live experiment results demonstrate significant improvements in student model performance while ensuring consistent and reliable generation of high-quality teacher labels from a continuous data stream of data.

24 February 2011 in Multimedia Tools and Applications
Verified
Journal Article
Closed access
PDF file
ZZZhe Zhao
Zhe Zhao
BCBin Cui
Bin Cui
GCGao Cong
Gao Cong
ZHZi Huang
Zi Huang
HSHeng Tao Shen
Heng Tao Shen

8 October 2024
Verified
Conference Paper
Open access
YWYi Wu
Yi Wu
DCDaryl Chang
Daryl Chang
JSJennifer She
Jennifer She
ZZZhe Zhao
Zhe Zhao
LWLi Wei
Li Wei
LHLukasz Heldt
Lukasz Heldt
We present the Learned Ranking Function (LRF), a system that takes short-term user-item behavior predictions as input and outputs a slate of recommendations that directly optimizes for long-term user satisfaction. Most previous work is based on optimizing the hyperparameters of a heuristic function. We propose to model the problem directly as a slate optimization problem with the objective of maximizing long-term user satisfaction. We also develop a novel constraint optimization algorithm that stabilizes objective tradeoffs for multi-objective optimization. We evaluate our approach with live experiments and describe its deployment on YouTube.

4 August 2023
Verified
Conference Paper
Open access
SIShibal Ibrahim
Shibal Ibrahim
WCWenyu Chen
Wenyu Chen
HHHussein Hazimeh
Hussein Hazimeh
NPNatalia Ponomareva
Natalia Ponomareva
ZZZhe Zhao
Zhe Zhao
RMRahul Mazumder
Rahul Mazumder
The sparse Mixture-of-Experts (Sparse-MoE) framework efficiently scales up model capacity in various domains, such as natural language processing and vision. Sparse-MoEs select a subset of the "experts" (thus, only a portion of the overall network) for each input sample using a sparse, trainable gate. Existing sparse gates are prone to convergence and performance issues when training with first-order optimization methods. In this paper, we introduce two improvements to current MoE approaches. First, we propose a new sparse gate: COMET, which relies on a novel tree-based mechanism. COMET is differentiable, can exploit sparsity to speed up computation, and outperforms state-of-the-art gates. Second, due to the challenging combinatorial nature of sparse expert selection, first-order methods are typically prone to low-quality solutions. To deal with this challenge, we propose a novel, permutation-based local search method that can complement first-order methods in training any sparse gate, e.g., Hash routing, Top-k, DSelect-k, and COMET. We show that local search can help networks escape bad initializations or solutions. We performed large-scale experiments on various domains, including recommender systems, vision, and natural language processing. On standard vision and recommender systems benchmarks, COMET+ (COMET with local search) achieves up to 13% improvement in ROC AUC over popular gates, e.g., Hash routing and Top-k, and up to 9% over prior differentiable gates e.g., DSelect-k. When Top-k and Hash gates are combined with local search, we see up to 100X reduction in the budget needed for hyperparameter tuning. Moreover, for language modeling, our approach improves over the state-of-the-art MoEBERT model for distilling BERT on 5/7 GLUE benchmarks as well as SQuAD dataset.

8 October 2024
Verified
Conference Paper
Open access
ZZZhen Zhang
Zhen Zhang
QLQingyun Liu
Qingyun Liu
YLYuening Li
Yuening Li
SBSourabh Bansod
Sourabh Bansod
MGMingyan Gao
Mingyan Gao
LLLiang Liu
Liang Liu
Multi-task prediction models and value models are the de-facto standard ranking components in modern large-scale content recommendation systems. However, they are typically optimized to model users’ passive consumption behaviors, and rank content in a way to grow only consumption-centric values. In this talk, we discuss the key insight that it is possible to model sparse participatory content-generation actions as well and grow ecosystem value through a new ranking system. We made the following key technical contributions in this system: (1) introducing ranking for content generation based on a categorization of user participation actions of different sparsity, including proxy intent action or access point clicks. (2) improving sparse task prediction quality and stability by causal task relationship modeling, conditional loss modeling and ResNet based shared bottom network. (3) personalizing the value model to minimize conflicts between different values, through e.g. ranking inspiring content higher for users who actively generate content. (4) conducting systematic evaluation of proposed approach in a large short-form video UGC (User-Generated Content) platform.