Uniform Diffusion
A discrete diffusion model in which, during the forward process, tokens transition uniformly across all categories. This uniform transition scheme simplifies the corruption process and provides a baseline for discrete diffusion methods, allowing tokens to be replaced by any category with equal probability at each step.
Timeline (last 90 days)
In the forward process, uniform diffusion jumps a token to any other category in the vocabulary at a uniform rate, ultimately tending toward a uniform distribution, and its reverse denoising is more difficult.
Traditional discrete diffusion (masked diffusion and uniform diffusion) completely erases token identity in a single corruption event, so the noisy state cannot express partial uncertainty (e.g., multiple candidates still being possible).
Under matched architectures and controlled training/sampling/tuning budgets, the symmetric form of binary diffusion consistently outperformed finely tuned masked diffusion and uniform diffusion baselines.
All Facts (3)
In the forward process, uniform diffusion jumps a token to any other category in the vocabulary at a uniform rate, ultimately tending toward a uniform distribution, and its reverse denoising is more difficult.
50%UnverifiedTraditional discrete diffusion (masked diffusion and uniform diffusion) completely erases token identity in a single corruption event, so the noisy state cannot express partial uncertainty (e.g., multiple candidates still being possible).
50%UnverifiedUnder matched architectures and controlled training/sampling/tuning budgets, the symmetric form of binary diffusion consistently outperformed finely tuned masked diffusion and uniform diffusion baselines.
50%