Unverified50% confidenceFactExact time
The vector field of ERFs is jointly determined by the forward and reverse rate matrices of the discrete diffusion model, and the continuous evolution rate of the residual depends on the row of the rate matrix corresponding to the current discrete state.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA recent paper proposed exponential residual flows (ERFs), which provide a deterministic ODE sampling form for discrete diffusion models and prove that it preserves the marginal distributions of the original diffusion model.70% similarUnverifiedThe sampling process of continuous diffusion models can be realized as a deterministic trajectory via the probability flow ordinary differential equation (ODE), supporting accelerated sampling and consistency distillation.69% similarUnverifiedThe arXiv:2610.09045v1 study proposes using conditional diffusion models to learn the transition kernel of time-homogeneous jump-diffusion processes in a data-driven manner and to generate new sample paths66% similarUnverifiedAutoregressive diffusion is a fusion variant built on diffusion models that introduces conditional dependence on historical time steps at each step of the reverse denoising process, causing the generation process to unfold along the time axis.66% similarUnverifiedFor a target distribution with tail index ν, the stable generative model can asymptotically correctly reproduce the scaling behavior of extreme quantiles if and only if α=ν.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115635API
curl https://kongchang.com/api/v1/knowledge/claims/1115635MCP
get_claim(id=1115635)