Research map / Image generation / Fast diffusion sampling
Variational Autoencoding Discrete Diffusion with Enhanced Dimensional Correlations Modeling
Tianyu Xie, Shuchen Xue, Zijin Feng, Tianyang Hu, Jiacheng Sun, Zhenguo Li, Cheng Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add latent variable and recognition model (VADD) to masked diffusion | MDLM · CIFAR-10, T=10 | FID: 334.3 -> 170.3 | Improved |
| add latent variable (VADD) | MDLM · CIFAR-10, T=100 | FID: 76.5 -> 50.5 | Improved |
| add latent variable (VADD) | MDLM · OpenWebText generative perplexity, T=16 | generative perplexity: 330.39 -> 194.08 | Improved |
| add latent variable (VADD) | MDLM · OpenWebText generative perplexity, T=1024 | generative perplexity: 104.58 -> 78.27 | Improved |
| add latent variable (VADD) | MDLM · LM1B | test perplexity: 27.70 -> 20.53 | Improved |
| add latent variable (VADD) | MDLM · OpenWebText | test perplexity: 23.07 -> 22.21 | Improved |
| add latent variable (VADD) | MDLM · binarized MNIST | BPD: 0.075 -> 0.063 | Improved |
| add latent variable (VADD) | MDLM · CIFAR-10 | BPD: 2.80 -> 2.74 | Improved |
| add latent variable (VADD) | MDLM · 2D toy checkerboard | JS-1: 1.395 -> 0.062 | Improved |
| add latent variable (VADD) | MDLM · zero-shot perplexity on six datasets | perplexity: Lambada 49.67 -> 47.30 | Improved |
| increase number of samples K in DELBO estimate (ablation) | VADD · OpenWebText | test perplexity: 22.56 -> 22.21 | Improved |
| recognition model takes only x0 instead of (x0, xt) (ablation) | VADD · text generation | posterior collapse | Hurt |
Closest papers by meaning
- Simplified and Generalized Masked Diffusion for Discrete Data (NeurIPS 2024 · similarity 0.89)
- DiffEnc: Variational Diffusion with a Learned Encoder (ICLR 2024 · similarity 0.88)
- Generalization in VAE and Diffusion Models: A Unified Information-Theoretic Analysis (ICLR 2025 · similarity 0.88)
- Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking (NeurIPS 2025 · similarity 0.88)
- DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents (ICML 2024 · similarity 0.88)
- Sparsely Supervised Diffusion (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
