English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The prevalent use of commercial and open-source diffusion models (DMs) fortext-to-image generation prompts risk mitigation to prevent undesiredbehaviors. Existing concept erasing methods in academia are all based on fullparameter or specification-based fine-tuning, from which we observe thefollowing issues: 1) Generation alternation towards erosion: Parameter driftduring target elimination causes alternations and potential deformations acrossall generations, even eroding other concepts at varying degrees, which is moreevident with multi-concept erased; 2) Transfer inability & deploymentinefficiency: Previous model-specific erasure impedes the flexible combinationof concepts and the training-free transfer towards other models, resulting inlinear cost growth as the deployment scenarios increase. To achievenon-invasive, precise, customizable, and transferable elimination, we groundour erasing framework on one-dimensional adapters to erase multiple conceptsfrom most DMs at once across versatile erasing applications. Theconcept-SemiPermeable structure is injected as a Membrane (SPM) into any DM tolearn targeted erasing, and meantime the alteration and erosion phenomenon iseffectively mitigated via a novel Latent Anchoring fine-tuning strategy. Onceobtained, SPMs can be flexibly combined and plug-and-play for other DMs withoutspecific re-tuning, enabling timely and efficient adaptation to diversescenarios. During generation, our Facilitated Transport mechanism dynamicallyregulates the permeability of each SPM to respond to different input prompts,further minimizing the impact on other concepts. Quantitative and qualitativeresults across ~40 concepts, 7 DMs and 4 erasing applications have demonstratedthe superior erasing of SPM. Our code and pre-tuned SPMs will be available onthe project page https://lyumengyao.github.io/projects/spm.

One-dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications