Research map / Image generation / Text-to-image diffusion
TeEFusion: Blending Text Embeddings to Distill Classifier-Free Guidance
Minghao Fu, Guo-Hua Wang, Xiaohao Chen, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang
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 DistillCFG guidance MLP with G(psi(w)) term (ablation) | TeEFusion (In-house T2I) · HPS Anime | HPS: 30.80 -> 31.05 | Improved |
| add F(c - null) fusion term to G(psi(w)) (ablation) | TeEFusion (In-house T2I) · HPS Anime | HPS: 31.05 -> 32.39 | Improved |
| add F(c - null) fusion term to G(psi(w)) (ablation) | TeEFusion (In-house T2I) · Anime prompts | ImageReward: 116.39 -> 128.50 | Improved |
| use TeEFusion instead of DistillCFG | SD3 student · HPS Anime | HPS: 31.14 -> 32.37 | Improved |
| use TeEFusion instead of DistillCFG | SD3 student · DPG-Bench | Overall: 84.13 -> 84.56 | Improved |
| use TeEFusion instead of DistillCFG | In-house T2I student · DPG-Bench | Overall: 83.52 -> 84.13 | Improved |
| distill W2SD+CFG teacher into 1x student | TeEFusion (SD3) · HPS Photo | HPS: 29.76 -> 29.84 | Improved |
| distill W2SD+CFG teacher into 1x student | TeEFusion (SD3) · DPG-Bench | Overall: 86.56 -> 84.56 | Hurt |
| distill from Euler+CFG teacher (ablation) | TeEFusion (In-house T2I) · Anime prompts | CLIP: 34.25 -> 33.58 | Hurt |
| distill from Euler+CFG teacher (ablation) | TeEFusion (In-house T2I) · Anime prompts | HPS: 30.65 -> 30.61 | No clear effect |
| distill from Z-Sampling+CFG teacher (ablation) | TeEFusion (In-house T2I) · Anime prompts | HPS: 31.99 -> 32.01 | No clear effect |
| distill from W2SD+CFG teacher (ablation) | TeEFusion (In-house T2I) · Anime prompts | PickScore: 22.49 -> 22.68 | Improved |
Closest papers by meaning
- A Cat Is A Cat (Not A Dog!): Unraveling Information Mix-ups in Text-to-Image Encoders through Causal Analysis and Embedding Optimization (NeurIPS 2024 · similarity 0.87)
- A Comprehensive Study of Decoder-Only LLMs for Text-to-Image Generation (CVPR 2025 · similarity 0.87)
- Training-Free Safe Text Embedding Guidance for Text-to-Image Diffusion Models (NeurIPS 2025 · similarity 0.87)
- Decoupled Textual Embeddings for Customized Image Generation (AAAI 2024 · similarity 0.87)
- Diffusion Adaptive Text Embedding for Text-to-Image Diffusion Models (NeurIPS 2025 · similarity 0.86)
- Würstchen: An Efficient Architecture for Large-Scale Text-to-Image Diffusion Models (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.
