Research map / Image generation / Autoregressive and latent generation
From Prediction to Perfection: Introducing Refinement to Autoregressive Image Generation
Cheng Cheng, Lin Song, Di An, Yicheng Xiao, Xuchong Zhang, Hongbin Sun, Ying Shan
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 TensorAR refinement to LlamaGen-B | LlamaGen-B · ImageNet 256 class-conditional | FID: 5.46 -> 4.71 | Improved |
| add TensorAR to LlamaGen-XXL | LlamaGen-XXL · ImageNet 384 class-conditional | FID: 2.34 -> 2.03 | Improved |
| add TensorAR to Open-MAGVIT2-B | Open-MAGVIT2-B · ImageNet 256 class-conditional | FID: 3.08 -> 2.91 | Improved |
| fine-tune baseline for same number of steps without TensorAR (ablation) | LlamaGen-B/L · ImageNet 256 class-conditional | FID | No clear effect |
| add TensorAR to LlamaGen for text-to-image | LlamaGen · GenEval overall | overall score: 0.32 -> 0.61 | Improved |
| add TensorAR to Janus-Pro-7B | Janus-Pro-7B · GenEval | overall score: 0.80 -> 0.83 overall; Counting 0.59 -> 0.53 | Mixed |
| add TensorAR to LlamaGen | LlamaGen · DPG-Bench | overall score: 43.13 -> 73.33 | Improved |
| add TensorAR to Janus-Pro-7B | Janus-Pro-7B · DPG-Bench | overall score: 84.19 -> 85.57 | Improved |
| use exponential noise schedule instead of linear/sine/square root (ablation) | TensorAR on LlamaGen-B · ImageNet 256 | FID: 4.79/4.75/4.84 -> 4.71 | Improved |
| increase window size k from 2 to 8 (ablation) | TensorAR on LlamaGen-B · ImageNet 256 | FID: 4.78 -> 4.68 | Improved |
| increase Q_in/Q_out depth from 1 to 4 (ablation) | TensorAR on LlamaGen-B · ImageNet 256 | FID: 4.71 -> 4.90 | Hurt |
Closest papers by meaning
- D-AR: Diffusion via Autoregressive Models (ICLR 2026 · similarity 0.91)
- ControlAR: Controllable Image Generation with Autoregressive Models (ICLR 2025 · similarity 0.91)
- HMAR: Efficient Hierarchical Masked Auto-Regressive Image Generation (CVPR 2025 · similarity 0.90)
- Hierarchical Masked Autoregressive Models with Low-Resolution Token Pivots (ICML 2025 · similarity 0.90)
- Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (NeurIPS 2024 · similarity 0.90)
- Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective (NeurIPS 2024 · similarity 0.89)
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.
