Research map / Image generation / Text-to-image diffusion
Improved and Accelerated Text-to-Image Generation With Collect, Reflect, and Refine
Shitong Shao, Zikai Zhou, Dian Xie, Yuetong Fang, Ye Tian, Lichen Bai, Bo Han, Zeke Xie
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 CoRe2 (weak model + W2S guidance) | SDXL · Pick-a-Pic | ImageReward: 0.5255 -> 0.7042 | Improved |
| add CoRe2 | SD3.5 · Pick-a-Pic | ImageReward: 0.8411 -> 0.8924 | Improved |
| add CoRe2 | SDXL · GenEval | overall: 54.39% -> 56.01% | Improved |
| add CoRe2 | SD3.5 · GenEval | overall: 69.71% -> 71.23% | Improved |
| add CoRe2 | SDXL · T2I-CompBench | overall: 0.4112 -> 0.4264 | Improved |
| add CoRe2 | SD3.5 · T2I-CompBench | overall: 0.5096 -> 0.5150 | Improved |
| add CoRe2 | LlamaGen · Pick-a-Pic | PickScore: 19.9200 -> 19.8600 | Hurt |
| add CoRe2 | LlamaGen · DrawBench | PickScore: 20.6200 -> 20.7800 | Improved |
| use Z-CoRe2 instead of Z-Sampling | SD3.5 · Pick-of-Pic | PickScore: 21.81 -> 22.11 | Improved |
| add CoRe2 on FLUX with schnell as weak model | FLUX · DrawBench | PickScore: 22.85 -> 22.88 | No clear effect |
| replace weak-model training: 0k iterations (use conditional output) versus 10k iterations (ablation) | CoRe2 weak model · SD3.5 | fast-mode performance | Improved |
| train weak model 50k iterations instead of 10k (ablation) | CoRe2 weak model · SD3.5 | full-capacity performance | Hurt |
Closest papers by meaning
- KOALA: Empirical Lessons Toward Memory-Efficient and Fast Diffusion Models for Text-to-Image Synthesis (NeurIPS 2024 · similarity 0.90)
- Image Regeneration: Evaluating Text-to-Image Model via Generating Identical Image with Multimodal Large Language Models (AAAI 2025 · similarity 0.89)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.89)
- R&B: Region and Boundary Aware Zero-shot Grounded Text-to-image Generation (ICLR 2024 · similarity 0.89)
- T2I-Adapter: Learning Adapters to Dig Out More Controllable Ability for Text-to-Image Diffusion Models (AAAI 2024 · similarity 0.89)
- SwiftBrush: One-Step Text-to-Image Diffusion Model with Variational Score Distillation (CVPR 2024 · 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.
