Research map / Image generation / Super-resolution
VARestorer: One-Step VAR Distillation for Real-World Image Super-Resolution
Yixuan Zhu, Shilin Ma, Haolin Wang, Ao Li, Yanzhe Jing, Yansong Tang, Lei Chen, Jiwen Lu, Jie Zhou
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 one-step VAR distillation instead of multi-step ControlNet VAR (ablation) | VARestorer · DIV2K-Val | MUSIQ: 62.22 -> 72.32 | Improved |
| add one-step VAR distillation instead of multi-step ControlNet VAR (ablation) | VARestorer · DIV2K-Val | CLIPIQA: 0.4794 -> 0.7669 | Improved |
| add cross-scale full attention instead of causal attention (ablation) | VARestorer · DIV2K-Val | MUSIQ: 63.72 -> 72.32 | Improved |
| add cross-scale full attention instead of causal attention (ablation) | VARestorer · DIV2K-Val | CLIPIQA: 0.3910 -> 0.7669 | Improved |
| add cross-scale full attention instead of causal attention (ablation) | VARestorer · DIV2K-Val | LPIPS: 0.4224 -> 0.3131 | Improved |
| add KL distribution matching loss (ablation) | VARestorer · DIV2K-Val | MUSIQ: 69.73 -> 72.32 | Improved |
| add KL distribution matching loss (ablation) | VARestorer · DIV2K-Val | CLIPIQA: 0.6682 -> 0.7669 | Improved |
| add KL distribution matching loss (ablation) | VARestorer · DIV2K-Val | NIQE: 4.372 -> 4.410 | Hurt |
| add BLIP text prompt at inference (ablation) | VARestorer · DIV2K-Val | MUSIQ: 67.05 -> 72.32 | Improved |
| use LoRA tuning instead of full fine-tuning (ablation) | VARestorer · DIV2K-Val | MUSIQ: 70.58 -> 72.32 | Improved |
| train at 1024 resolution instead of 512 (ablation) | VARestorer · DIV2K-Val | MUSIQ: 72.32 -> 72.11 | No clear effect |
| train longer (20K and 25K steps vs 10K) (ablation) | VARestorer · DIV2K-Val | all metrics | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SeeSR-50 — SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution (CVPR 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Visual Autoregressive Modeling for Image Super-Resolution (ICML 2025 · similarity 0.92)
- Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (NeurIPS 2024 · similarity 0.91)
- NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution (NeurIPS 2026 · similarity 0.91)
- HMAR: Efficient Hierarchical Masked Auto-Regressive Image Generation (CVPR 2025 · similarity 0.89)
- TSD-SR: One-Step Diffusion with Target Score Distillation for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.88)
- Fundamental Limits of Visual Autoregressive Transformers: Universal Approximation Abilities (ICML 2025 · similarity 0.88)
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.
