Xuan Zhu
Research Subjects
Xuan Zhu has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Andreas Hochwagen
Department of Biology, New York University
1 shared publication - Chih-Hsien Hsia
Department of Computer Science and Information Engineering, National Ilan University, Yilan 260007, Taiwan
1 shared publication Haidong Sun
1 shared publication- Jiayu Li
Syracuse University
1 shared publication - Luis Andre Vale-Silva
Cellzome GmbH, GlaxoSmithKline, Heidelberg, Germany
1 shared publication - Nancy M. Hollingsworth
Department of Biochemistry and Cell Biology, Stony Brook University, Stony Brook, NY 11794, USA
1 shared publication
Latest Publications
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 13291-13292
Multi-Turn Reinforcement Learning for Large Language Models: From Theory to Practice with Amazon SageMaker AI
Production large language model (LLM) based systems such as coding agents, web navigators, and tool-calling assistants operate over multiple turns of interaction with users, tools, and environments. Pretrained LLMs, depending on their size, can either underperform in these settings due to misalignment with the system's interaction mechanics, or, when capable, incur prohibitive latency. Fine-tuning right-sized models addresses both accuracy and latency, but training such multi-turn agents requires Reinforcement Learning (RL), where the model acts as a policy optimizing long-horizon outcomes across sequential interactions. This poses challenges absent from single-turn settings: credit assignment over long trajectories, reward design for sparse and delayed feedback, state and context management as observation histories grow, environment scaling for parallel rollout collection, and training stability under prompt/environment distribution shift. This hands-on problem-solving tutorial provides both a rigorous algorithmic and practical introduction to multi-turn RL finetuning for LLMs. Using Amazon SageMaker AI, participants progress through four labs: (1) environment and reward function design, (2) multi-turn trajectory collection and Group Relative Policy Optimization (GRPO)-based training, (3) reward densification and credit assignment strategies, and (4) evaluation, failure diagnosis and deployment. We cover state-of-the-art multi-turn RL finetuning algorithms, turn-level vs. trajectory-level reward design, and production grade monitoring for reward hacking detection. The tutorial targets machine learning (ML) engineers, data scientists, and researchers who build agentic LLM systems. No prior RL experience is required. All materials will be publicly available on GitHub.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 13291-13292
Open AccessMulti-Turn Reinforcement Learning for Large Language Models: From Theory to Practice with Amazon SageMaker AI
Production large language model (LLM) based systems such as coding agents, web navigators, and tool-calling assistants operate over multiple turns of interaction with users, tools, and environments. Pretrained LLMs, depending on their size, can either underperform in these settings due to misalignment with the system's interaction mechanics, or, when capable, incur prohibitive latency. Fine-tuning right-sized models addresses both accuracy and latency, but training such multi-turn agents requires Reinforcement Learning (RL), where the model acts as a policy optimizing long-horizon outcomes across sequential interactions. This poses challenges absent from single-turn settings: credit assignment over long trajectories, reward design for sparse and delayed feedback, state and context management as observation histories grow, environment scaling for parallel rollout collection, and training stability under prompt/environment distribution shift. This hands-on problem-solving tutorial provides both a rigorous algorithmic and practical introduction to multi-turn RL finetuning for LLMs. Using Amazon SageMaker AI, participants progress through four labs: (1) environment and reward function design, (2) multi-turn trajectory collection and Group Relative Policy Optimization (GRPO)-based training, (3) reward densification and credit assignment strategies, and (4) evaluation, failure diagnosis and deployment. We cover state-of-the-art multi-turn RL finetuning algorithms, turn-level vs. trajectory-level reward design, and production grade monitoring for reward hacking detection. The tutorial targets machine learning (ML) engineers, data scientists, and researchers who build agentic LLM systems. No prior RL experience is required. All materials will be publicly available on GitHub.
- 28 December 2023
- Published by Mason Publish Group in Electrical & Electronic Engineering Research
- Vol. 3 (1)
Research on university entrepreneurship education in the context of “mass entrepreneurship and innovation”
The 3rd International Conference on Electrical and Electronic Engineering, Design, Communication and Artificial Intelligence in 2023 (EEDCAI 2023) took place in December26-28, 2023 in Beijing China. It is organized by Beijing Society of Ecology and Agriculture Institute, co-organized by Wuhan University of Technology,Zhongke (Urban and Rural Areas) Planning and Design and Research Institute,China Engineering Technology Innovation Center Co-organizer,Shanghai University…
- 28 December 2023
- Published by Mason Publish Group in Electrical & Electronic Engineering Research
- Vol. 3 (1)
Research on university entrepreneurship education in the context of “mass entrepreneurship and innovation”
The 3rd International Conference on Electrical and Electronic Engineering, Design, Communication and Artificial Intelligence in 2023 (EEDCAI 2023) took place in December26-28, 2023 in Beijing China. It is organized by Beijing Society of Ecology and Agriculture Institute, co-organized by Wuhan University of Technology,Zhongke (Urban and Rural Areas) Planning and Design and Research Institute,China Engineering Technology Innovation Center Co-organizer,Shanghai University…
- Open Access
- 27 February 2019
- Published by Springer Nature in Nature Communications
- Vol. 10 (1) , 1-15
Persistent DNA-break potential near telomeres increases initiation of meiotic recombination on short chromosomes
Faithful meiotic chromosome inheritance and fertility rely on the stimulation of meiotic crossover recombination by potentially genotoxic DNA double-strand breaks (DSBs). To avoid excessive damage, feedback mechanisms down-regulate DSBs, likely in response to initiation of crossover repair. In Saccharomyces cerevisiae, this regulation requires the removal of the conserved DSB-promoting protein Hop1/HORMAD during chromosome synapsis. Here, we identify privileged end-adjacent regions (EARs) spanning roughly 100 kb near all telomeres that escape DSB down-regulation. These regions retain Hop1 and continue to break in pachynema despite normal synaptonemal complex deposition. Differential retention of Hop1 requires the disassemblase Pch2/TRIP13, which preferentially removes Hop1 from telomere-distant sequences, and is modulated by the histone deacetylase Sir2 and the nucleoporin Nup2. Importantly, the uniform size of EARs among chromosomes contributes to disproportionately high DSB and repair signals on short chromosomes in pachynema, suggesting that EARs partially underlie the curiously high recombination rate of short chromosomes.
- 27 February 2019
- Published by Springer Nature in Nature Communications
- Vol. 10 (1) , 1-15
Open AccessPersistent DNA-break potential near telomeres increases initiation of meiotic recombination on short chromosomes
Faithful meiotic chromosome inheritance and fertility rely on the stimulation of meiotic crossover recombination by potentially genotoxic DNA double-strand breaks (DSBs). To avoid excessive damage, feedback mechanisms down-regulate DSBs, likely in response to initiation of crossover repair. In Saccharomyces cerevisiae, this regulation requires the removal of the conserved DSB-promoting protein Hop1/HORMAD during chromosome synapsis. Here, we identify privileged end-adjacent regions (EARs) spanning roughly 100 kb near all telomeres that escape DSB down-regulation. These regions retain Hop1 and continue to break in pachynema despite normal synaptonemal complex deposition. Differential retention of Hop1 requires the disassemblase Pch2/TRIP13, which preferentially removes Hop1 from telomere-distant sequences, and is modulated by the histone deacetylase Sir2 and the nucleoporin Nup2. Importantly, the uniform size of EARs among chromosomes contributes to disproportionately high DSB and repair signals on short chromosomes in pachynema, suggesting that EARs partially underlie the curiously high recombination rate of short chromosomes.