Chengen Huang
Chengen Huang is a researcher at Alibaba Group whose papers span language models, mobile systems, and research agents. He studied as an undergraduate at Claremont Graduate University and as an MS student at Tsinghua University, and presented Trinity, a mobile-emulation project, at OSDI in 2022. He co-authored technical reports on Yi-Lightning in 2024 and Qwen3 in 2025. His later work evaluated DeepResearch reports, explored training language models to generate research ideas, and developed reinforcement learning for search-agent abstention.
Timeline
-
2015–2023
Claremont, Tsinghua and Trinity
He studies as an undergraduate at Claremont Graduate University from 2015 to 2019 and as an MS student at Tsinghua University from 2020 to 2023. At OSDI in 2022, he presents Trinity: High-Performance Mobile Emulation through Graphics Projection.
-
December 2024
The Yi-Lightning technical report
In December 2024, he coauthors the Yi-Lightning technical report, which describes a Mixture-of-Experts language model with optimized KV caching.
-
May 2025
The Qwen3 technical report
In May 2025, he coauthors the Qwen3 technical report, which describes dense and Mixture-of-Experts language models spanning 0.6 to 235 billion parameters.
-
October 2025
Evaluating DeepResearch reports
In October 2025, he coauthors Understanding DeepResearch via Reports, which introduces a 100-query benchmark for comparing report quality, redundancy, and factuality.
-
February 2026
Training models for idea generation
In February 2026, he coauthors DeepInnovator, a training framework designed to help large language models generate research ideas.
-
July 2026
Training search agents to abstain
In July 2026, he coauthors To Answer or to Abstain, which introduces Abstention-Aware Reinforcement Learning (AWA-RL) to shape abstention rewards using query-specific capabilities and on-policy training observations.
-
Since 2025
Researcher at Alibaba Group
He is a researcher at Alibaba Group.
Career
- Since 2025AlibabaResearcher
Education
- 2023Tsinghua UniversityMS student
- 2019Claremont Graduate UniversityUndergraduate student
Most-cited works
- Qwen3 Technical ReportWith Baosong Yang, Beichen Zhang, Binyuan Hui, Fei Huang, Le Yu and Rui Men
- Yi: Open Foundation Models by 01.AI
- REMOTE SENSING OF CROP GROWTH CHARACTERISTICS IN GREENHOUSES
- DeepInnovator: Triggering the Innovative Capabilities of LLMs
- Land Cover Change Community-based Processing and Analysis System (LC-ComPS): Lessons Learned from Technology Infusion
Sources and credits
Photo: We have no published photo of Chengen Huang yet. Send us a free-licensed photo.
- Chengen Huang — Lacuna
- Chengen Huang | USENIX
- Yi-Lightning Technical Report
- Qwen3 Technical Report
- Understanding DeepResearch via Reports
- DeepInnovator: Triggering the Innovative Capabilities of LLMs
- To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning
