Research map / Image generation / Text-to-image diffusion
ImageRAGTurbo: Towards One-step Text-to-Image Generation with Retrieval-Augmented Diffusion Models
Peijie Qiu, Hariharan Ramshankar, Arnau Ramisa, Amit Kumar K C, René Vidal, Vamsi Salaka, Rahul Bhagat
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 retrieval augmentation with H-space adapter | SD Turbo v2-1 (1-step) · TIFA | TIFA: 0.779 -> 0.801 | Improved |
| add retrieval augmentation with H-space adapter | SD Turbo v2-1 (1-step) · MS-COCO | FID: 26.04 -> 25.59 | Improved |
| add retrieval augmentation with H-space adapter | SD Turbo v2-1 (1-step) · MS-COCO | CLIP score: 0.319 -> 0.323 | Improved |
| add retrieval augmentation | SD Turbo v2-1 (1-step) · TIFA | AES: 5.85 -> 5.88 | Mixed |
| training-free direct H-space injection with fixed blend w=0.8 (ablation) | SD Turbo · TIFA | TIFA: 0.779 -> 0.781 | No clear effect |
| per-prompt brute-force search of blend strength in direct injection (ablation) | SD Turbo · TIFA | TIFA: 0.781 -> 0.816 | Improved |
| use trained H-space adapter instead of retrieval-conditioned RDM | retrieval-augmented generation · TIFA | TIFA: 0.725 -> 0.801 | Improved |
| one-step ImageRAGTurbo versus 50-step SD v2-1 | few-step distillation · TIFA | TIFA: 0.811 -> 0.801 | Hurt |
| one-step ImageRAGTurbo versus 4-step LCM | few-step distillation · TIFA | TIFA: 0.764 -> 0.801 | Improved |
| add H-space adapter (36M params) and retrieval | SD Turbo · latency | ms/image: 113.8 -> 116.7 | No clear effect |
Closest papers by meaning
- InitNO: Boosting Text-to-Image Diffusion Models via Initial Noise Optimization (CVPR 2024 · similarity 0.91)
- On Discrete Prompt Optimization for Diffusion Models (ICML 2024 · similarity 0.90)
- PaRa: Personalizing Text-to-Image Diffusion via Parameter Rank Reduction (ICLR 2025 · similarity 0.90)
- Get What You Want, Not What You Don't: Image Content Suppression for Text-to-Image Diffusion Models (ICLR 2024 · similarity 0.90)
- OneActor: Consistent Subject Generation via Cluster-Conditioned Guidance (NeurIPS 2024 · similarity 0.90)
- Minority-Focused Text-to-Image Generation via Prompt Optimization (CVPR 2025 · similarity 0.90)
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.
