Zijing Hu

B2-DiffuRL. [CVPR 25] A framework named B^2-DiffuRL for RL-based diffusion model fine-tuning.

AsynDM. [ICLR 26] Asynchronous diffusion models allocate individual pixels with varying timestep schedules, yielding improved text-to-image alignment.

D-Fusion. [ICML 25] Denoising trajectory fusion, a method to construct RL-trainable visually consistent samples.

Awesome-Multimodal-RL. This repository collects research papers on reinforcement learning in generative multimodal models.

hu-zijing.github.io. Hi, I'm Zijing Hu, a Ph.D student of Zhejiang University.