Chenrui Ma 马晨睿

Research Interests

My research centers on efficient generative modeling, trustworthy machine learning, and multimodal AI.

News

Selected Publications (All Publications; * indicates equal contribution)

Citation counts are from Semantic Scholar; Stars are from GitHub. Last checked September 25, 2026 (UTC).

Chenrui Ma, Xi Xiao, Lin Zhao, Tianyang Wang, Ferdinando Fioretto, Yanning Shen

Drift Flow Matching

NeurIPS 2026 Main Track [Paper]

0 Citations

Enables both one-step generation and multi-step refinement in drift models, trading inference compute for sample quality.

---

Chenrui Ma, Xi Xiao, Tianyang Wang, Xiao Wang, Yanning Shen

Stochastic Interpolants via Conditional Dependent Coupling

NeurIPS 2026 Main Track [Paper] [Code]

6 Citations 1 Star

Enables end-to-end, coarse-to-fine image generation in pixel space through conditional dependent coupling and a single shared model.

---

Chenrui Ma, Xi Xiao, Tianyang Wang, Xiao Wang, Yanning Shen

Learning Straight Flows: Variational Flow Matching for Efficient Generation

CVPR 2026 Main Conference [Paper] [Code]

9 Citations 6 Stars

Uses variational latent conditioning and a straightness objective to learn straighter flows for efficient one-step generation.

---

Chenrui Ma, Xi Xiao, Tianyang Wang, Xiao Wang, Yanning Shen

CAD-VAE: Leveraging Correlation-Aware Latents for Comprehensive Fair Disentanglement

AAAI 2026 Main Track [Paper] [Code]

10 Citations 2 Stars

Models correlations between target and sensitive attributes with a shared latent code for fair representation learning and image editing.

---

Xi Xiao*, Chenrui Ma*, Yunbei Zhang*, Chen Liu, Zhuxuanzi Wang, Yanshu Li, Lin Zhao, Guosheng Hu, Tianyang Wang, Hao Xu

Not All Directions Matter: Towards Structured and Task-Aware Low-Rank Model Adaptation

ACL 2026 Main Conference [Paper]

14 Citations

Filters task-irrelevant LoRA directions and coordinates updates across layers, with no additional inference cost.

---

Xinru Jiang*, Lin Zhao*, Xi Xiao, Yunbei Zhang, Janet Wang, Chenrui Ma, Haolin Li, Yanzhi Wang, Yifan Gong, Octavia Camps

Dynamic Hub-and-Spoke Memory for Streaming Video Understanding

EMNLP 2026 Findings [Paper]

Combines question-relevant textual memory with recent visual frames for streaming video question answering, without additional training.

---

Chenrui Ma

Transition Flow Matching

arXiv [Paper]

1 Citation

Learns direct transitions along generative trajectories, enabling one-step and few-step sampling with flexible step sizes.

Awards & Honors

• [2026] Computer Science Scholar Fellowship, University of Virginia.
• [2025] AAAI-2026 Scholarship, Singapore.
• [2024] University-level Second Prize Scholarship (Top 20%), Central South University.
• [2023] University-level Third Prize Scholarship (Top 30%), Central South University.
• [2023] 21st China Undergraduate Mathematical Competition in Modeling, National Second Prize, Top 15% nationwide.
• [2023] 4th National College Student Mathematical Modeling Competition, National First Prize, Top 5% nationwide.

Service & Activities