Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.18150 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910232208211968 |
|---|---|
| author | Sun, Mengyu Yang, Ziyuan Zhou, Zunlong Liu, Junxu Hu, Haibo Zhang, Yi |
| author_facet | Sun, Mengyu Yang, Ziyuan Zhou, Zunlong Liu, Junxu Hu, Haibo Zhang, Yi |
| contents | Diffusion models (DMs) are widely used for text-to-image generation, but their strong generative capabilities also raise concerns about unsafe or undesirable content. Concept erasure aims to mitigate these risks by removing specific concepts from pretrained models. However, recent studies show that such methods often suppress rather than fully eliminate target concepts, leaving models vulnerable to awakening attacks. Existing approaches primarily rely on white-box access through optimization or inversion, while concept awakening under black-box constraints remains underexplored. In this work, we revisit the denoising process from a trajectory perspective and show that concept erasure mainly disrupts early-stage text-semantic alignment but does not fully prevent semantic information from propagating along the denoising dynamics. As generation proceeds, the model increasingly depends on the evolving noisy state rather than textual conditions, which creates an opportunity to bypass erased mappings. Motivated by this observation, we propose ConceptAgent, a training-free, black-box, multi-agent framework that awakens erased concepts by initializing the denoising trajectory from surrogate-guided noisy states. Extensive experiments demonstrate that ConceptAgent enables accurate and controllable awakening of erased concepts under black-box settings without access to model parameters, gradients, or internal representations. These results highlight fundamental limitations of current concept erasure methods and provide new insights into the dynamic nature of semantic control in DMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_18150 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Whispers in the Noise: Surrogate-Guided Concept Awakening via a Multi-Agent Framework Sun, Mengyu Yang, Ziyuan Zhou, Zunlong Liu, Junxu Hu, Haibo Zhang, Yi Artificial Intelligence Diffusion models (DMs) are widely used for text-to-image generation, but their strong generative capabilities also raise concerns about unsafe or undesirable content. Concept erasure aims to mitigate these risks by removing specific concepts from pretrained models. However, recent studies show that such methods often suppress rather than fully eliminate target concepts, leaving models vulnerable to awakening attacks. Existing approaches primarily rely on white-box access through optimization or inversion, while concept awakening under black-box constraints remains underexplored. In this work, we revisit the denoising process from a trajectory perspective and show that concept erasure mainly disrupts early-stage text-semantic alignment but does not fully prevent semantic information from propagating along the denoising dynamics. As generation proceeds, the model increasingly depends on the evolving noisy state rather than textual conditions, which creates an opportunity to bypass erased mappings. Motivated by this observation, we propose ConceptAgent, a training-free, black-box, multi-agent framework that awakens erased concepts by initializing the denoising trajectory from surrogate-guided noisy states. Extensive experiments demonstrate that ConceptAgent enables accurate and controllable awakening of erased concepts under black-box settings without access to model parameters, gradients, or internal representations. These results highlight fundamental limitations of current concept erasure methods and provide new insights into the dynamic nature of semantic control in DMs. |
| title | Whispers in the Noise: Surrogate-Guided Concept Awakening via a Multi-Agent Framework |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.18150 |