AI RESEARCH
Neural Continuous-Time Markov Chain: Discrete Diffusion via Decoupled Jump Timing and Direction
arXiv CS.LG
•
ArXi:2604.15694v1 Announce Type: new Discrete diffusion models based on continuous-time Marko chains (CTMCs) have shown strong performance on language and discrete data generation, yet existing approaches typically parameterize the reverse rate matrix as a single object -- via concrete scores, clean-data predictions ($x_0$-parameterization), or denoising distributions -- rather than aligning the parameterization with the intrinsic CTMC decomposition into jump timing and jump direction.