Research map / Image generation / Autoregressive and latent generation
Auto-Regressive Image Synthesis with Integrated Quantization
Fangneng Zhan, Yingchen Yu, Rongliang Wu, Jiahui Zhang, Kaiwen Cui, Changgong Zhang, Shijian Lu
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 |
|---|---|---|---|
| replace VQ-GAN with IQ-VAE without regularization (ablation) | VQ-GAN · ADE20K | FID: 35.50 -> 31.88 | Improved |
| replace VQ-GAN with IQ-VAE without regularization (ablation) | VQ-GAN · ADE20K | SWD: 21.50 -> 19.14 | Improved |
| replace VQ-GAN with IQ-VAE without regularization (ablation) | VQ-GAN · ADE20K | LPIPS: 0.421 -> 0.441 | Improved |
| add variational regularizer (ablation) | IQ-VAE(None) · ADE20K | FID: 31.88 -> 31.41 | Improved |
| add variational regularizer (ablation) | IQ-VAE(None) · ADE20K | SWD: 19.14 -> 18.71 | Improved |
| add variational regularizer (ablation) | IQ-VAE(None) · ADE20K | LPIPS: 0.441 -> 0.450 | Improved |
| add Gumbel sampling (ablation) | IQ-VAE(VR) · ADE20K | FID: 31.41 -> 29.77 | Improved |
| add Gumbel sampling (ablation) | IQ-VAE(VR) · ADE20K | SWD: 18.71 -> 17.44 | Improved |
| add Gumbel sampling (ablation) | IQ-VAE(VR) · ADE20K | LPIPS: 0.450 -> 0.447 | No clear effect |
| add Gumbel sampling (applied every 4 iterations) (ablation) | IQ-VAE · training speed | iteration/s: 3.0 -> 2.8 | Hurt |
| replace VQ-GAN with full IQ-VAE | VQ-GAN · DeepFashion | FID: 16.49 -> 11.15 | Improved |
| replace VQ-GAN with full IQ-VAE | VQ-GAN · CelebA-HQ (Edge) | FID: 16.23 -> 14.71 | Improved |
Closest papers by meaning
- Regularized Vector Quantization for Tokenized Image Synthesis (CVPR 2023 · similarity 0.89)
- Efficient Generative Modeling with Residual Vector Quantization-Based Tokens (ICML 2025 · similarity 0.89)
- Boosting Generative Image Modeling via Joint Image-Feature Synthesis (NeurIPS 2025 · similarity 0.89)
- A Simple Approach to Unifying Diffusion-based Conditional Generation (ICLR 2025 · similarity 0.88)
- CycleVAR: Repurposing Autoregressive Model for Unsupervised One-Step Image Translation (ICCV 2025 · similarity 0.88)
- Distilled Decoding 2: One-step Sampling of Image Auto-regressive Models with Conditional Score Distillation (NeurIPS 2025 · 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.
