TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866917535179341824 |
|---|---|
| author | Cho, Hyunmin Ahn, Donghoon Hong, Susung Kim, Jee Eun Kim, Seungryong Jin, Kyong Hwan |
| author_facet | Cho, Hyunmin Ahn, Donghoon Hong, Susung Kim, Jee Eun Kim, Seungryong Jin, Kyong Hwan |
| contents | Diffusion models achieve state-of-the-art image generation but often produce semantic inconsistencies, or hallucinations. Existing inference-time guidance methods rely on external signals or architectural modifications, adding computational overhead. We propose $\mathbf{T}$angential $\mathbf{A}$mplifying $\mathbf{G}$uidance $\mathbf{(TAG)}$, a training-free, architecture-agnostic, plug-and-play guidance method that operates purely on trajectory signals. TAG uses an intermediate sample as a projection basis and amplifies the tangential components of the estimated score to correct the sampling trajectory. A first-order Taylor analysis shows that this steers the state toward higher-probability regions of the data manifold, reducing inconsistencies and improving fidelity while adding negligible overhead to existing samplers. Code is available at our Project Page (https://hyeon-cho.github.io/TAG/). |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_04533 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling Cho, Hyunmin Ahn, Donghoon Hong, Susung Kim, Jee Eun Kim, Seungryong Jin, Kyong Hwan Computer Vision and Pattern Recognition Diffusion models achieve state-of-the-art image generation but often produce semantic inconsistencies, or hallucinations. Existing inference-time guidance methods rely on external signals or architectural modifications, adding computational overhead. We propose $\mathbf{T}$angential $\mathbf{A}$mplifying $\mathbf{G}$uidance $\mathbf{(TAG)}$, a training-free, architecture-agnostic, plug-and-play guidance method that operates purely on trajectory signals. TAG uses an intermediate sample as a projection basis and amplifies the tangential components of the estimated score to correct the sampling trajectory. A first-order Taylor analysis shows that this steers the state toward higher-probability regions of the data manifold, reducing inconsistencies and improving fidelity while adding negligible overhead to existing samplers. Code is available at our Project Page (https://hyeon-cho.github.io/TAG/). |
| title | TAG: Tangential Amplifying Guidance for Hallucination-Resistant Sampling |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2510.04533 |