Stochastic Self-Guidance for Training-Free Enhancement of Diffusion Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915832591810560 |
|---|---|
| author | Chen, Chubin Zhu, Jiashu Feng, Xiaokun Huang, Nisha Zhu, Chen Wu, Meiqi Mao, Fangyuan Wu, Jiahong Chu, Xiangxiang Li, Xiu |
| author_facet | Chen, Chubin Zhu, Jiashu Feng, Xiaokun Huang, Nisha Zhu, Chen Wu, Meiqi Mao, Fangyuan Wu, Jiahong Chu, Xiangxiang Li, Xiu |
| contents | Classifier-free Guidance (CFG) is a widely used technique in modern diffusion models for enhancing sample quality and prompt adherence. However, through an empirical analysis on Gaussian mixture modeling with a closed-form solution, we observe a discrepancy between the suboptimal results produced by CFG and the ground truth. The model's excessive reliance on these suboptimal predictions often leads to semantic incoherence and low-quality outputs. To address this issue, we first empirically demonstrate that the model's suboptimal predictions can be effectively refined using sub-networks of the model itself. Building on this insight, we propose S$^2$-Guidance, a novel method that leverages stochastic block-dropping during the forward process to construct stochastic sub-networks, effectively guiding the model away from potential low-quality predictions and toward high-quality outputs. Extensive qualitative and quantitative experiments on text-to-image and text-to-video generation tasks demonstrate that S$^2$-Guidance delivers superior performance, consistently surpassing CFG and other advanced guidance strategies. Our code will be released. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_12880 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Stochastic Self-Guidance for Training-Free Enhancement of Diffusion Models Chen, Chubin Zhu, Jiashu Feng, Xiaokun Huang, Nisha Zhu, Chen Wu, Meiqi Mao, Fangyuan Wu, Jiahong Chu, Xiangxiang Li, Xiu Computer Vision and Pattern Recognition Classifier-free Guidance (CFG) is a widely used technique in modern diffusion models for enhancing sample quality and prompt adherence. However, through an empirical analysis on Gaussian mixture modeling with a closed-form solution, we observe a discrepancy between the suboptimal results produced by CFG and the ground truth. The model's excessive reliance on these suboptimal predictions often leads to semantic incoherence and low-quality outputs. To address this issue, we first empirically demonstrate that the model's suboptimal predictions can be effectively refined using sub-networks of the model itself. Building on this insight, we propose S$^2$-Guidance, a novel method that leverages stochastic block-dropping during the forward process to construct stochastic sub-networks, effectively guiding the model away from potential low-quality predictions and toward high-quality outputs. Extensive qualitative and quantitative experiments on text-to-image and text-to-video generation tasks demonstrate that S$^2$-Guidance delivers superior performance, consistently surpassing CFG and other advanced guidance strategies. Our code will be released. |
| title | Stochastic Self-Guidance for Training-Free Enhancement of Diffusion Models |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2508.12880 |