LZ

Lei ZANG臧雷

LLM Algorithm Expert, Ant Group

Ant Group — RL post-training of Ling-series LLMs for LingGuang

Master's Supervisor, Xiamen University

Xiamen University — Supervisor of Master of Applied Statistics

Email: zanglei.zl@antgroup.com

About

I am Lei ZANG (臧雷), an LLM Algorithm Expert at Ant Group, working on RL post-training of Ling-series large language models for LingGuang. Since 2023 I have also served as a Master's Supervisor of Applied Statistics at Xiamen University. My research interests include LLM post-training, reinforcement learning, and long-horizon agents.

Profiles: ORCID 0009-0005-8869-6034 · Google Scholar · GitHub @drop-out

Positions

Ant Group — LLM Algorithm Expert 2019 — Present
Responsible for RL post-training of Ling-series LLMs for LingGuang.
Xiamen University — Master's Supervisor, Applied Statistics 2023 — Present

Education

Xiamen University — Master of Finance 2019
Wang Yanan Institute for Studies in Economics (WISE), Xiamen University. Advisor: Prof. Wei Zhong (钟威). Thesis: Volatility Indicator Construction via Deep Learning and the Application of Multiple Volatility Indicators.
Fordham University — Master of Science in Finance 2018

Publications

ACL 2024
Chengyue Yu*†, Lei Zang*†, Jiaotuan Wang, Chenyi Zhuang, Jinjie Gu
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, Volume 3: System Demonstrations (ACL 2024)
Token-free, character-by-character generation for precise format control (format accuracy > 0.96); open-sourced on ModelScope; demo video on YouTube.
EMNLP 2026
Zongyue Li*, Chengyue Yu*, Lei Zang*, Chenyi Zhuang, Linjian Mo, Leilei Gan†
Main Conference of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026)

* equal contribution; † corresponding author. Full records on ORCID and Google Scholar.

Honors & Awards

2017 CCF BDCI (CCF Big Data & Computing Intelligence Contest) 🏆 Champion · 1st-Place Solution
2017 IJCAI Data Mining Contest 🥉 3rd-Place Solution
2018 China Collegiate Computing Contest — Big Data Challenge 🏆 Champion · 1st-Place Solution
The winning approach used an RNN many-to-many architecture with step-by-step forecasting — predicting each future step sequentially — which is strikingly similar to today's next-token prediction paradigm of large language models.
2019 National Outstanding Finance Master's Thesis Nomination
  • 6th National Outstanding Finance Master's Thesis Award (第六届全国优秀金融硕士学位论文), among 149 theses nationwide — 30 winners + 15 nominations.
  • Thesis: Volatility Indicator Construction via Deep Learning and the Application of Multiple Volatility Indicators.
  • Coverage: WISE, Xiamen University news report