Haoran Xu
- Decoder-Hybrid-Decoder Architecture for Efficient Reasoning with Long Generation
- Information-Theoretic Reward Decomposition for Generalizable RLHF
- Spike4DGS: Towards High-Speed Dynamic Scene Rendering with 4D Gaussian Splatting via a Spike Camera Array
- Uni-RL: Unifying Online and Offline RL via Implicit Value Regularization