Research map / Image generation / Text-to-image diffusion
DreamLite: A Lightweight On-Device Unified Model for Image Generation and Editing
Kailai Feng, Yuxiang WEI, Bo Chen, yang pan, Hu Ye, Songwei Liu, Chenqian Yan, Yuan Gao, Wangmeng Zuo
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 |
|---|---|---|---|
| in-context spatial concatenation instead of Pix2Pix channel concat (T2I->Edit) (ablation) | DreamLite · ImgEdit | score: 3.67 -> 3.88 | Improved |
| in-context instead of Pix2Pix at unified stage (ablation) | DreamLite · GenEval | score: 0.61 -> 0.71 | Improved |
| add edit pretraining stage before unified training (TPJ) (ablation) | DreamLite · ImgEdit | score: 3.14 -> 3.94 | Improved |
| add edit pretraining stage before unified training (TPJ) (ablation) | DreamLite · GenEval | score: 0.65 -> 0.71 | Improved |
| in-context T2I training with blank image (ablation) | DreamLite · GenEval | score: 0.70 -> 0.65 | Hurt |
| add RLHF (ReFL) after TPJ pretraining (ablation) | DreamLite · ImgEdit | score: 3.94 -> 4.11 | Improved |
| add RLHF (ReFL) after TPJ pretraining (ablation) | DreamLite · GenEval | score: 0.71 -> 0.72 | Improved |
| add 4-step DMD distillation (ablation) | DreamLite · GenEval | score: 0.72 -> 0.70 | Hurt |
| add 4-step DMD distillation (ablation) | DreamLite · ImgEdit | score: 4.11 -> 3.8 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- SnapGen (0.38B) — SnapGen: Taming High-Resolution Text-to-Image Models for Mobile Devices with Efficient Architectures and Training (CVPR 2025) · better on 4 of 4 shared results
- SANA-0.6B — SANA: Efficient High-Resolution Text-to-Image Synthesis with Linear Diffusion Transformers (ICLR 2025) · better on 3 of 3 shared results
- OmniGen2 (4B) — OmniGen: Unified Image Generation (CVPR 2025) · better on 1 of 1 shared result
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
- SnapGen: Taming High-Resolution Text-to-Image Models for Mobile Devices with Efficient Architectures and Training (CVPR 2025 · similarity 0.89)
- DreamOmni: Unified Image Generation and Editing (CVPR 2025 · similarity 0.89)
- Upcycling Text-to-Image Diffusion Models for Multi-Task Capabilities (ICML 2025 · similarity 0.88)
- EditAR: Unified Conditional Generation with Autoregressive Models (CVPR 2025 · similarity 0.87)
- DreamSteerer: Enhancing Source Image Conditioned Editability using Personalized Diffusion Models (NeurIPS 2024 · similarity 0.86)
- PixArt-$\alpha$: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis (ICLR 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.
