Zhengkai Jiang

My research spans image and video generation, diffusion models, multi-modal learning, and visual representation learning.

At Tencent, I focus on post-training the Hunyuan multi-modal foundation models — aligning large-scale generative models with human preference.

Reward Modeling & RL for Vision Generation — aligning text-to-image and text-to-video diffusion models with human preference. I build reward models that capture what people actually value in a generated image or video, then post-train generative policies against them with GRPO, NFT, OPD and ReFL.