Li Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Li Wei has a total of 57 co-authors. Here are the top co-authors by number of shared publications:
- Lukasz Heldt
YouTube, Google, Mountain View, CA, USA
6 shared publications - Aniruddh Nath
Google, Mountain View, USA
5 shared publications - Lichan Hong
Google, Mountain View, CA, USA
5 shared publications - Nikhil Khani
Google, Mountain View, USA
4 shared publications - Daryl Chang
YouTube, Google, Mountain View, CA, USA
3 shared publications - Raghunandan Keshavan
YouTube, Google, Mountain View, USA
3 shared publications
Latest Publications
- Open Access
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 1462-1470
Self-Evolving Recommendation System: End-To-End Autonomous Model Optimization With LLM Agents
Optimizing large-scale machine learning systems, such as recommendation models for global video platforms, requires navigating a massive hyperparameter search space and, more critically, designing sophisticated optimizers, architectures, and reward functions to capture nuanced user behaviors. Achieving substantial improvements in these areas is a non-trivial task, traditionally relying on extensive manual iterations to test new hypotheses. We propose a self-evolving system that leverages Large Language Models (LLMs), specifically those from Google’s Gemini family, to autonomously generate, train, and deploy high-performing, complex model changes within an end-to-end automated workflow. The self-evolving system consists of an Offline Agent (Fast Loop) that performs high-throughput hypothesis generation to optimize for proxy metrics, and an Online Agent (Slow Loop) that validates candidates against delayed north star business metrics in live production. Our agents act as specialized Machine Learning Engineers (MLEs): they exhibit deep reasoning capabilities, discovering novel improvements in optimization algorithms and model architecture, and formulating innovative reward functions that target long-term user engagement. The effectiveness of this approach is demonstrated through several successful production launches at YouTube, confirming that autonomous, LLM-driven evolution can surpass traditional engineering workflows in both development velocity and model performance.
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 1462-1470
Open AccessSelf-Evolving Recommendation System: End-To-End Autonomous Model Optimization With LLM Agents
Optimizing large-scale machine learning systems, such as recommendation models for global video platforms, requires navigating a massive hyperparameter search space and, more critically, designing sophisticated optimizers, architectures, and reward functions to capture nuanced user behaviors. Achieving substantial improvements in these areas is a non-trivial task, traditionally relying on extensive manual iterations to test new hypotheses. We propose a self-evolving system that leverages Large Language Models (LLMs), specifically those from Google’s Gemini family, to autonomously generate, train, and deploy high-performing, complex model changes within an end-to-end automated workflow. The self-evolving system consists of an Offline Agent (Fast Loop) that performs high-throughput hypothesis generation to optimize for proxy metrics, and an Online Agent (Slow Loop) that validates candidates against delayed north star business metrics in live production. Our agents act as specialized Machine Learning Engineers (MLEs): they exhibit deep reasoning capabilities, discovering novel improvements in optimization algorithms and model architecture, and formulating innovative reward functions that target long-term user engagement. The effectiveness of this approach is demonstrated through several successful production launches at YouTube, confirming that autonomous, LLM-driven evolution can surpass traditional engineering workflows in both development velocity and model performance.
- Open Access
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 1534-1541
Token Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models
Large Recommendation Models (LRMs) have demonstrated promising capabilities in industry-scale recommendation tasks. However, holistically integrating traditional signals into these transformer-based architectures effectively and efficiently remains a major challenge. Conventional approaches that "textualize" these signals directly or create discrete item representations often lead to excessively long prompts, substantial memory footprints, and high computational overhead. To overcome these limitations, we propose "Token Factory", a framework designed to transform traditional signals into "soft tokens" that can be directly processed by LRMs. This approach enables efficient integration and compression of heterogeneous input features, preventing prompt length explosion while enhancing model performance. We detail the architecture of Token Factory and present experimental results validating its effectiveness in a production-scale recommendation environment.
- 27 September 2026
- Published by Association for Computing Machinery (ACM)
- p. 1534-1541
Open AccessToken Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models
Large Recommendation Models (LRMs) have demonstrated promising capabilities in industry-scale recommendation tasks. However, holistically integrating traditional signals into these transformer-based architectures effectively and efficiently remains a major challenge. Conventional approaches that "textualize" these signals directly or create discrete item representations often lead to excessively long prompts, substantial memory footprints, and high computational overhead. To overcome these limitations, we propose "Token Factory", a framework designed to transform traditional signals into "soft tokens" that can be directly processed by LRMs. This approach enables efficient integration and compression of heterogeneous input features, preventing prompt length explosion while enhancing model performance. We detail the architecture of Token Factory and present experimental results validating its effectiveness in a production-scale recommendation environment.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7456-7464
Addressing Intent Dissonance in Cross-Domain Distillation on YouTube
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 7456-7464
Open AccessAddressing Intent Dissonance in Cross-Domain Distillation on YouTube
- Open Access
- 28 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 711-715
Learning to Alleviate Familiarity Bias in Video Recommendation
- 28 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 711-715
Open AccessLearning to Alleviate Familiarity Bias in Video Recommendation
- Open Access
- 22 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 1122-1125
Zero-shot Cross-domain Knowledge Distillation: A Case study on YouTube Music
- 22 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 1122-1125
Open AccessZero-shot Cross-domain Knowledge Distillation: A Case study on YouTube Music
- Open Access
- 22 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 958-961
Enhancing Online Ranking Systems via Multi-Surface Co-Training for Content Understanding
- 22 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 958-961
Open AccessEnhancing Online Ranking Systems via Multi-Surface Co-Training for Content Understanding
- Open Access
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 758-761
Bridging the Gap: Unpacking the Hidden Challenges in Knowledge Distillation for Online Ranking Systems
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.
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 758-761
Open AccessBridging the Gap: Unpacking the Hidden Challenges in Knowledge Distillation for Online Ranking Systems
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.
- Open Access
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1004-1009
Learned Ranking Function: From Short-term Behavior Predictions to Long-term User Satisfaction
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.
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1004-1009
Open AccessLearned Ranking Function: From Short-term Behavior Predictions to Long-term User Satisfaction
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.
- Top Cited PapersOpen Access
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1039-1044
Better Generalization with Semantic IDs: A Case Study in Ranking for Recommendations
Randomly-hashed item ids are used ubiquitously in recommendation models. However, the learned representations from random hashing prevents generalization across similar items, causing problems of learning unseen and long-tail items, especially when item corpus is large, power-law distributed, and evolving dynamically. In this paper, we propose using content-derived features as a replacement for random ids. We show that simply replacing ID features with content-based embeddings can cause a drop in quality due to reduced memorization capability. To strike a good balance of memorization and generalization, we propose to use Semantic IDs [15], a compact and discrete item representation, as a replacement for random item ids. Semantic IDs are learned from frozen content embeddings using RQ-VAE and thus can capture the hierarchy of concepts in items. Similar to content embeddings, the compactness of Semantic IDs poses a problem of adaption in recommendation models. We propose novel methods for adapting Semantic IDs in industry-scale ranking models, through hashing sub-pieces of of the Semantic-ID sequences. In particular, we find that the SentencePiece model [10] that is commonly used in LLM tokenization outperforms manually crafted pieces such as N-grams. To the end, we evaluate our approaches in a real-world ranking model for YouTube recommendations. Our experiments demonstrate that Semantic IDs can replace the direct use of video IDs by improving the generalization ability on new and long-tail item slices without sacrificing overall model quality.
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 1039-1044
Top Cited PapersOpen AccessBetter Generalization with Semantic IDs: A Case Study in Ranking for Recommendations
Randomly-hashed item ids are used ubiquitously in recommendation models. However, the learned representations from random hashing prevents generalization across similar items, causing problems of learning unseen and long-tail items, especially when item corpus is large, power-law distributed, and evolving dynamically. In this paper, we propose using content-derived features as a replacement for random ids. We show that simply replacing ID features with content-based embeddings can cause a drop in quality due to reduced memorization capability. To strike a good balance of memorization and generalization, we propose to use Semantic IDs [15], a compact and discrete item representation, as a replacement for random item ids. Semantic IDs are learned from frozen content embeddings using RQ-VAE and thus can capture the hierarchy of concepts in items. Similar to content embeddings, the compactness of Semantic IDs poses a problem of adaption in recommendation models. We propose novel methods for adapting Semantic IDs in industry-scale ranking models, through hashing sub-pieces of of the Semantic-ID sequences. In particular, we find that the SentencePiece model [10] that is commonly used in LLM tokenization outperforms manually crafted pieces such as N-grams. To the end, we evaluate our approaches in a real-world ranking model for YouTube recommendations. Our experiments demonstrate that Semantic IDs can replace the direct use of video IDs by improving the generalization ability on new and long-tail item slices without sacrificing overall model quality.
- Open Access
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 4882-4893
Improving Training Stability for Multitask Ranking Models in Recommender Systems
Recommender systems play an important role in many content platforms. While most recommendation research is dedicated to designing better models to improve user experience, we found that research on stabilizing the training for such models is severely under-explored. As recommendation models become larger and more sophisticated, they are more susceptible to training instability issues, i.e., loss divergence, which can make the model unusable, waste significant resources and block model developments. In this paper, we share our findings and best practices we learned for improving the training stability of a real-world multitask ranking model for YouTube recommendations. We show some properties of the model that lead to unstable training and conjecture on the causes. Furthermore, based on our observations of training dynamics near the point of training instability, we hypothesize why existing solutions would fail, and propose a new algorithm to mitigate the limitations of existing solutions. Our experiments on YouTube production dataset show the proposed algorithm can significantly improve training stability while not compromising convergence, comparing with several commonly used baseline methods.
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 4882-4893
Open AccessImproving Training Stability for Multitask Ranking Models in Recommender Systems
Recommender systems play an important role in many content platforms. While most recommendation research is dedicated to designing better models to improve user experience, we found that research on stabilizing the training for such models is severely under-explored. As recommendation models become larger and more sophisticated, they are more susceptible to training instability issues, i.e., loss divergence, which can make the model unusable, waste significant resources and block model developments. In this paper, we share our findings and best practices we learned for improving the training stability of a real-world multitask ranking model for YouTube recommendations. We show some properties of the model that lead to unstable training and conjecture on the causes. Furthermore, based on our observations of training dynamics near the point of training instability, we hypothesize why existing solutions would fail, and propose a new algorithm to mitigate the limitations of existing solutions. Our experiments on YouTube production dataset show the proposed algorithm can significantly improve training stability while not compromising convergence, comparing with several commonly used baseline methods.