Reinforcement learning & verification
Jiacheng Xu 徐嘉诚
PhD Student, College of Computing and Data Science (CCDS)
Nanyang Technological University, Singapore
AI Researcher, Skywork AI
Learning from feedback. Building better ways to verify.
Email: jiacheng005@e.ntu.edu.sg
Google Scholar
LinkedIn
GitHub
Hugging Face
Biography
I am a PhD student at the College of Computing and Data Science (CCDS), Nanyang Technological University, and an AI researcher at Skywork AI. My research focuses on reinforcement learning and verification for code language models, particularly executable feedback, test generation, and inference-time selection.
Research Interests
- Reinforcement Learning
- Code Language Models
- Test Generation and Executable Verification
- Inference-Time Scaling and Selection
NOPO ↗
Jiacheng Xu, Shuo He, Fuxiang Zhang, Chaojie Wang, and Bo An.
NeurIPS 2026. (Spotlight, ~0.95% of submissions) [Project] [Paper] [PDF]
Learn from reliable failures when passing a check does not certify correctness.
WLTS
Backtracking with Linear-in-Depth Search-Space Growth: Width-Limited Tree Search for Large Language Models
Jiacheng Xu, Fuxiang Zhang, Chaojie Wang, and Bo An.
NeurIPS 2026. (Spotlight, ~0.95% of submissions)
ERPO ↗
Jiacheng Xu, Feng Chen, Xiuneng Xu, and Bo An.
EMNLP 2026. (Oral, ~2.6% of submissions) [Project] [Paper] [Code] [PDF]
Adapt code models at test time using execution agreement, without oracle outputs.
TCS ↗
Jiacheng Xu, Wentao Zhang, Zhiyi Lyu, Fuxiang Zhang, Chaojie Wang, Yang Liu, and Bo An.
Findings of EMNLP 2026. [Project] [Paper] [Code] [Models] [PDF]
Learn sound, adversarial tests and use them to select better programs.
Self-Verification ↗
Fuxiang Zhang, Jiacheng Xu, Chaojie Wang, Ce Cui, Yang Liu, and Bo An.
NeurIPS 2025. [Paper] [Code] [PDF]
FTD ↗
Chao Chen*, Jiacheng Xu*, Weijian Liao, Hao Ding, Zongzhang Zhang, Yang Yu, and Rui Zhao.
AAAI 2024. [Project] [Paper] [Code] [PDF]
ILI ↗
Jiacheng Xu, Chao Chen, Fuxiang Zhang, Lei Yuan, Zongzhang Zhang, and Yang Yu.
KDD 2023. [Project] [Paper] [Code] [PDF]
Turn symbolic rules into rewards that guide visual reinforcement learning.
MADAC ↗
Ke Xue*, Jiacheng Xu*, Lei Yuan, Miqing Li, Chao Qian, Zongzhang Zhang, and Yang Yu.
NeurIPS 2022. (Spotlight, ~3.4% of submissions) [Paper] [Code] [PDF]
* Equal Contribution
Awards
- National Scholarship, 2023