Research map / Image generation / Super-resolution
Mixture of Ranks with Degradation-Aware Routing for One-Step Real-World Image Super-Resolution
Xiao He, Zhijun Tu, Kun Cheng, Mingrui Zhu, Jie Hu, Nannan Wang, Xinbo Gao
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 dense LoRA with LoRA+MoE (ablation) | one-step Real-ISR (SD 2.1) · Real-world SR ablation (matches DRealSR) | CLIPIQA: 0.670 -> 0.689 | Improved |
| treat each LoRA rank as an expert (MoR-v1) instead of LoRA+MoE (ablation) | one-step Real-ISR (SD 2.1) · Real-world SR ablation (matches DRealSR) | CLIPIQA: 0.689 -> 0.704 | Improved |
| add zero experts with standard load-balancing loss (ablation) | MoR-v1 · Real-world SR ablation (matches DRealSR) | MANIQA: 0.491 -> 0.479 | Hurt |
| add degradation-aware load-balancing loss together with zero experts (ablation) | MoR · Real-world SR ablation (matches DRealSR) | TRES: 79.41 -> 81.78 | Improved |
| use MoR-DASR instead of S3Diff | S3Diff · DIV2K-Val (one-step Real-ISR) | LPIPS: 0.258 -> 0.289 | Hurt |
| use MoR-DASR instead of OSEDiff | OSEDiff · RealSR (one-step Real-ISR) | MANIQA: 0.468 -> 0.512 | Improved |
| use MoR-DASR instead of OSEDiff | OSEDiff · DRealSR (one-step Real-ISR) | PSNR: 28.29 -> 28.37 | No clear effect |
Closest papers by meaning
- See More Details: Efficient Image Super-Resolution by Experts Mining (ICML 2024 · similarity 0.88)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.87)
- Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration (NeurIPS 2026 · similarity 0.87)
- Multilinear Mixture of Experts: Scalable Expert Specialization through Factorization (NeurIPS 2024 · similarity 0.87)
- Complexity Experts are Task-Discriminative Learners for Any Image Restoration (CVPR 2025 · similarity 0.87)
- A Provably Effective Method for Pruning Experts in Fine-tuned Sparse Mixture-of-Experts (ICML 2024 · similarity 0.86)
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.
