Research map / Image generation / Text-to-image diffusion
TCFG: Truncated Classifier-Free Guidance for Efficient and Scalable Text-to-Image Acceleration
Xiaomeng Fu, Jia Li
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 TCFG truncation of text-conditional computation | DDIM-50 on SD 1.5 · MS-COCO val | speedup / FID: 1x -> 1.70x; FID 65.18 -> 66.43 | Improved |
| TCFG vs DDIM-35 at comparable NFE | SD 1.5 · MS-COCO val | FID: 77.86 -> 66.43 | Improved |
| TCFG vs DeepCache at same speedup | SD 1.5 · MS-COCO val | FID: 82.16 -> 66.43 | Improved |
| TCFG vs TokenMerge | SD 1.5 · MS-COCO val | FID: 101.52 -> 66.43 | Improved |
| add guidance rescheduling (warmup, decay) with truncation | DDIM-50 · MS-COCO val | HPSv2: 28.31 -> 28.97 | Improved |
| add guidance rescheduling and truncation | DDIM-50 · MS-COCO val | LAION-Aes: 5.25 -> 5.42 | Improved |
| add TCFG to ControlNet with control branch truncation | ControlNet · edge-to-image | speedup: x1.72 | Improved |
| add warmup stage (tw below 1000) (ablation) | TCFG · SD 1.5 ablation | convergence / alignment | Mixed |
| increase scaled learning rate C (ablation) | TCFG · SD 1.5 ablation | semantic convergence / MUSIQ | Mixed |
| use DPM-Solver-35 instead of DDIM-35 as efficient sampler | SD 1.5 · MS-COCO val | speedup: 1.33x -> 0.76x | Hurt |
Later papers that compared with this method
1 later accepted paper reports TCFG as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling (ICML 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DeepCache — DeepCache: Accelerating Diffusion Models for Free (CVPR 2024) · better on 6 of 7 shared results
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
- CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.90)
- Rethinking Training for De-biasing Text-to-Image Generation: Unlocking the Potential of Stable Diffusion (CVPR 2025 · similarity 0.90)
- Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models (AAAI 2025 · similarity 0.90)
- Towards Better Alignment: Training Diffusion Models with Reinforcement Learning Against Sparse Rewards (CVPR 2025 · similarity 0.89)
- Denoising Autoregressive Transformers for Scalable Text-to-Image Generation (ICLR 2025 · similarity 0.89)
- AdaDiff: Adaptive Step Selection for Fast Diffusion Models (AAAI 2025 · similarity 0.89)
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.
