Skip to content

[new-research] arXiv:2604.15829 — Beyond Text Prompts: Precise Concept Erasure through Text-Image Collaboration #78

@github-actions

Description

@github-actions

New arXiv Research Paper

arXiv ID: 2604.15829
Title: Beyond Text Prompts: Precise Concept Erasure through Text-Image Collaboration
Authors: Jun Li, Lizhi Xiong, Ziqiang Li, Weiwei Jiang, Zhangjie Fu, Yong Li, Guo-Sen Xie
Published: 2026-04-17T08:32:39Z
URL: https://arxiv.org/abs/2604.15829

Abstract (first 300 chars)

Text-to-image generative models have achieved impressive fidelity and diversity, but can inadvertently produce unsafe or undesirable content due to implicit biases embedded in large-scale training datasets. Existing concept erasure methods, whether text-only or image-assisted, face trade-offs: textu…

Suggested OWASP Mapping

No automatic mapping — manual review recommended

Suggested Action

Review this paper against the GenAI Security Crosswalk entries and update relevant mapping files if new attack patterns or mitigations are identified.

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions