About me

I am a PhD student in the Department of Computing at Imperial College London, learning to train energy-based models under the supervision of Yingzhen Li. I completed my undergraduate studies in the School of Computer Science and Engineering at Sun Yat-sen University and previously worked as a research intern at:

• Apple MLR (06/2024 - 09/2024): large-scale video diffusion models training
• Shell AI (09/2023 - 12/2023): flow matching for petrol station floorplan generation
• Tencent AI Lab (07/2021 - 09/2022): discrete energy-based modelling for set function learning (NeurIPS 2022 Oral)
• Tencent Jarvis Lab (05/2020 - 06/2021): representation learning with variational inference (ACL 2021 Oral)

My PhD research is deep generative modelling, and currently focusing on diffusion post-training and test-time scaling, as well as few-step generation and distillation:

• Test-time scaling and post-training for diffusion models [paper]
• Few-step generation with optimal covariance matching [paper]; one-step distillation [paper1, paper2]
• Neural flow sampler for unnormalized distributions [paper1, paper2]
• Training energy-based models without MCMC [paper1, paper2]

Publications

* denotes equal contribution; check the full list here

• Discrete Neural Flow Samplers with Locally Equivariant Transformer
  Zijing Ou, Ruixiang Zhang, and Yingzhen Li. Neural Processing Information Systems (NeurIPS), 2025. [paper] [code]

• Improving Probabilistic Diffusion Models With Optimal Diagonal Covariance Matching
  Oral (Accept rate~1.8%)
  Zijing Ou*, Mingtian Zhang*, Andi Zhang, Tim Xiao, Yingzhen Li, and David Barber. International Conference on Learning Representations (ICLR), 2025. [paper] [code]

• Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces
  Tobias Schröder*, Zijing Ou*, Yingzhen Li, and Andrew Duncan. Neural Processing Information Systems (NeurIPS), 2024. [paper] [code]

• Energy Discrepancies: A Score-Independent Loss for Energy-Based Models
  Tobias Schröder, Zijing Ou, Jen Ning Lim, Yingzhen Li, Sebastian Vollmer, and Andrew Duncan. Neural Processing Information Systems (NeurIPS), 2023. [paper] [code]

• Learning Neural Set Functions Under the Optimal Subset Oracle
  Oral (Accept rate~1.7%)
  Zijing Ou, Tingyang Xu, Qinliang Su, Yingzhen Li, Peilin Zhao, and Yatao Bian. Neural Processing Information Systems (NeurIPS), 2022. [paper] [code] [slides] [poster]

Services

• Conference Reviewing: ICLR (2024-2026), ICML (2022-2025), NeurIPS (2022-2025), AISTAT (2023-2026), IJCAI (2022-2024), ACL (2022), NAACL (2022)