Research map / Image generation / Text-to-image diffusion
Adaptive Routing of Text-to-Image Generation Requests Between Large Cloud Model and Light-Weight Edge Model
Zewei Xin, Qinya Li, Chaoyue Niu, Fan Wu, Guihai Chen
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 multi-metric quality optimization (ablation) | RouteT2I · COCO2014 win rate (40% routing rate) | normalized win rate improvement: 27.37 -> 30.60 | Improved |
| add token selection gate (ablation) | RouteT2I · COCO2014 win rate (80% routing rate) | normalized win rate improvement: 19.24 -> 21.94 | Improved |
| add dual-gate MoE (ablation) | RouteT2I · COCO2014 win rate (50% routing rate) | normalized win rate improvement: 22.09 -> 25.81 | Improved |
| add dual-gate MoE (ablation) | RouteT2I · COCO2014 win rate (80% routing rate) | normalized win rate improvement: 21.62 -> 21.94 | Improved |
| use RouteT2I instead of ZOOTER router | routing model · COCO2014 (50% routing rate) | relative performance improvement DeltaP: 77.95 -> 83.97 | Improved |
| use RouteT2I instead of random routing for cost saving | routing · COCO2014 | cost saving at 50% DeltaP: 70.24 | Improved |
| use RouteT2I instead of ZOOTER router | routing model · LAION | relative performance improvement DeltaP: 76.45 -> 81.38 | Improved |
| use RouteT2I instead of random routing | routing · COCO2014 HPSv2 | DeltaP on HPSv2: 50.00 -> 69.50 | Improved |
| use RouteT2I instead of random routing | routing · COCO2014 MPS | DeltaP on MPS: 50.00 -> 57.57 | Improved |
Closest papers by meaning
- Policy Optimized Text-to-Image Pipeline Design (NeurIPS 2025 · similarity 0.84)
- VL-RouterBench: A Benchmark for Vision-Language Model Routing (CVPR 2026 · similarity 0.83)
- DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models (NeurIPS 2026 · similarity 0.83)
- T2ISafety: Benchmark for Assessing Fairness, Toxicity, and Privacy in Image Generation (CVPR 2025 · similarity 0.83)
- SnapGen: Taming High-Resolution Text-to-Image Models for Mobile Devices with Efficient Architectures and Training (CVPR 2025 · similarity 0.83)
- EC-DIT: Scaling Diffusion Transformers with Adaptive Expert-Choice Routing (ICLR 2025 · similarity 0.83)
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.
