Scholar
Sheng Cheng
Google Scholar ID: TWAwdYsAAAAJ
Arizona State University
Citations & Impact
Citations
195
H-index
8
i10-index
6
Publications
14
Co-authors
7
list available
Publications
Resume (English only)
Academic Achievements
- - Published 'Latent Space Energy-based Neural ODEs' in TMLR 2025
- - Published 'TripletCLIP: Improving Compositional Reasoning of CLIP via Vision-Language Negatives' in NeurIPS 2024
- - Published 'Precision or Recall? An Analysis of Image Captions for Training Text-to-Image Generation Model' in EMNLP 2024, Findings
- - Published 'Revising Text-to-Image Prior for Improved Text Conditioned Image Generations' in CVPR 2024
- - Published 'WOUAF: Weight Modulation for User Attribution and Fingerprinting in Text-to-Image Diffusion Models' in CVPR 2024
- - Published 'Self-supervised Learning to Discover Physical Objects and Predict Their Interactions from Raw Videos' in NeurIPS AI4Science workshop, 2023
- - Published 'Adversarial Bayesian Augmentation for Single-Source Domain Generalization' in ICCV 2023
- - Published 'SSR-GNNs: Stroke-based Sketch Representation with Graph Neural Networks' in CVPR Sketch workshop, 2022
- - Published 'Data-Driven Learning of Three-Point Correlation Functions as Microstructure Representations' in Acta Materialia, 2022
- - Published 'Evaluating the Robustness of Bayesian Neural Networks Against Different Types of Attacks' in CVPR Adversarial Machine Learning workshop, 2021
Research Experience
- - Amazon AGI, Feb 2025 - Present, working on image/video generation
- - Amazon AWS, Fall 2024, mitigating hallucinations in MLLMs using hard negative samples
- - Bosch US, Summer 2024, masked controlled autonomous driving
Education
- - Ph.D., Arizona State University, Advisor: Yezhou Yang
- - M.Eng., University of Illinois at Urbana-Champaign, Advisor: Ruoyu Sun
- - B.S., Huazhong University of Science and Technology
Background
- Research interests include image/video generation, multimodal LLM, vision & language (particularly in Text-to-Image generation), domain generalization & robustness, and AI in Science.
