Scholar
Zecheng He
Google Scholar ID: tcwZh8oAAAAJ
Meta GenAI
Citations & Impact
Citations
1,827
H-index
17
i10-index
19
Publications
20
Co-authors
0
Publications
Resume
Academic Achievements
- - Selected Publications:
- - Movie Weaver: Tuning-Free Multi-Concept Video Personalization with Anchored Prompts, CVPR 2025
- - Movie Gen: A Cast of Media Foundation Models, Technical Report, 2024
- - Imagine Yourself: Tuning-Free Personalized Image Generation, arXiv Preprint, 2024
- - MoCha: Towards Movie-Grade Talking Character Synthesis, arXiv Preprint, 2024
- - DirectorLLM for Human-Centric Video Generation, arXiv Preprint, 2024
- - Token-Shuffle: Towards High-Resolution Image Generation with Autoregressive Models, arXiv Preprint, 2024
- - Trainable Projected Gradient Method for Robust Fine-tuning, CVPR 2023
- - Adversarial Medical Image with Hierarchical Feature Hiding, IEEE Transactions on Medical Imaging, 2023
- - ActionBert: Leveraging User Actions for Semantic Understanding of User Interfaces, AAAI 2021
- - New Models for Understanding and Reasoning about Speculative Execution Attacks, HPCA 2021
- - Sensitive-sample Fingerprinting of Deep Neural Networks, CVPR 2019
- - Model Inversion Attacks Against Collaborative Inference, ACSAC 2019
Research Experience
- - Staff research scientist at Meta GenAI, working on image and video generation
- - Project Experience:
- - Core contributor to MovieGen, leading personalized video generation pretrain and post-train
- - Lead and first author of Imagine Yourself, Meta's first personalized image generation model
- - Core contributor to Llama native image generation, focusing on native model architecture design and development
Education
- - Ph.D. from Princeton University (Advisor information not provided)
Background
- - Research Interests: Generative models for image and video, including diffusion-based and native generation models
- - Professional Field: Generative models, image and video generation
- - Brief Introduction: Staff research scientist at Meta GenAI, working on image and video generation, including personalized image/video diffusion models, native image generation, and controllable generation.