Two new papers propose machine unlearning methods for MLLMs

Arxiv CS.CV (computer vision)

SPACE introduces source-free concept erasure for MLLMs. Visual-Noise Guided In-Context Distillation offers an alternative unlearning approach using noisy visual prompts.

mllmmultimodal

See 1 more source

SPACE: Source-free Proxy Anchor Concept Erasure for MLLMs23 hours agoZhijing Zhang, Jiaqi Ding, Qianshan Wei, Nan Zhou, Jiaqi Li, Yongliang Wu, Tongxin Zhu, Xiaolin Fang

Arxiv CS.CV (computer vision)·Junkai ChenYuhao HeJunxiang YouRuiqi LiuChenyu WangShu Wu

1 day ago