Language Modeling on WikiText zero-shot transfer (test)
33.22PerplexityAbsorbing Uniform Diffusion
Evaluation Results
| Method | Links | |
|---|---|---|
| Absorbing Uniform DiffusionLoss=ELBO, Param.=Denoiser, Zero-shot transfer=true2026.05 | 33.22 | |
| Masked DiffusionLoss=ELBO, Param.=Denoiser, Zero-shot transfer=true2026.05 | 35.16 | |
| Uniform DiffusionLoss=Cross-Entropy, Param.=LOO-Denoiser, Zero-shot transfer=true2026.05 | 36.67 | |
| Uniform DiffusionLoss=ELBO, Param.=LOO-Denoiser, Zero-shot transfer=true2026.05 | 37.23 | |
| Uniform DiffusionLoss=ELBO, Param.=Denoiser, Zero-shot transfer=true2026.05 | 37.44 | |
| Uniform DiffusionLoss=Cross-Entropy, Param.=Denoiser, Zero-shot transfer=true2026.05 | 39.17 | |
| Max Coupling Uniform DiffusionLoss=ELBO, Param.=LOO-Denoiser, Zero-shot transfer=true2026.05 | 49.41 | |
| Max Coupling Uniform DiffusionLoss=ELBO, Param.=Denoiser, Zero-shot transfer=true2026.05 | 51.4 |