Research map / Image generation / Fast diffusion sampling
Improved Autoregressive Modeling with Distribution Smoothing
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Smooth the data distribution and denoise with a conditional autoregressive model | PixelCNN++ · CIFAR-10 | FID: 54.25 -> 29.83 | Helped |
| Smooth the data distribution and denoise with a conditional autoregressive model | PixelCNN++ · CIFAR-10 | Inception: 5.30 -> 7.84 ± .07 | Helped |
| Smooth the data distribution and denoise with a conditional autoregressive model | PixelCNN++ · CIFAR-10 | KID: 0.046 -> 0.022 | Helped |
| Smooth the data distribution and denoise with a conditional autoregressive model | PixelCNN++ · CIFAR-10 | BPD: 2.92 -> ≤ 3.53 | Hurt |
| Use gradient-based single-step denoising rather than conditional autoregressive denoising (ablation) | Smoothed PixelCNN++ · CIFAR-10 | FID: 29.83 -> 57.53 | Hurt |
| Use gradient-based single-step denoising rather than conditional autoregressive denoising (ablation) | Smoothed PixelCNN++ · CIFAR-10 | Inception: 7.84 ± .07 -> 7.50± .08 | Hurt |
| Use gradient-based single-step denoising rather than conditional autoregressive denoising (ablation) | Smoothed PixelCNN++ · CIFAR-10 | KID: 0.022 -> 0.052 | Hurt |
| Increase mixture components in the denoising model from two to three or four (ablation) | MADE smoothing model · Olympics | sample quality | Helped |
| Train the smoothing variance by jointly optimizing ELBO instead of selecting it heuristically (ablation) | PixelCNN++ smoothing model · CIFAR-10 | sample quality | Hurt |
| Apply single-step gradient denoising without smoothing the training data (ablation) | PixelCNN++ · CIFAR-10 | sample quality | No clear effect |
| Use randomized smoothing with three mixture components instead of direct MADE with six components | MADE · Rings | negative log-likelihood: 2.81 -> 2.71 | Helped |
| Use randomized smoothing with three mixture components instead of direct MADE with six components | MADE · Olympics | negative log-likelihood: 0.80 -> 0.80 | No clear effect |
Closest papers by meaning
- Autoregressive Image Generation without Vector Quantization (NeurIPS 2024 · similarity 0.87)
- Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective (NeurIPS 2024 · similarity 0.86)
- FlexVAR: Flexible Visual Autoregressive Modeling without Residual Prediction (NeurIPS 2025 · similarity 0.86)
- Distilled Decoding 1: One-step Sampling of Image Auto-regressive Models with Flow Matching (ICLR 2025 · similarity 0.86)
- Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (NeurIPS 2024 · similarity 0.85)
- Adaptive Median Smoothing: Adversarial Defense for Unlearned Text-to-Image Diffusion Models at Inference Time (ICML 2025 · similarity 0.85)
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.