[2604.03956] VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models

Computer Science > Computer Vision and Pattern Recognition arXiv:2604.03956 (cs) [Submitted on 5 Apr 2026] Title:VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models Authors:Ravi Ranjan, Agoritsa Polyzou View a PDF of the paper titled VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models, by Ravi Ranjan and Agoritsa Polyzou View PDF HTML (experimental) Abstract:Vision-language-action (VLA) models are emerging as embodied foundation models for robotic manipulation, but their deployment introduces a new unlearning challenge: removing unsafe, spurious, or privacy-sensitive behaviors without degrading perception, language grounding, and action control. In OpenVLA-style policies, behavior is produced through a fused visual encoder, a cross-modal projector, and a language backbone that predicts tokenized robot actions, so undesirable knowledge can be distributed across perception, alignment, and reasoning/action layers rather than confined to a single module. Consequently, partial unlearning applied only to the vision stack or only to the language backbone is often insufficient, while conventional unlearning baselines designed for standalone vision or language models may leave residual forgetting or incur unnecessary utility loss in embodied settings. We propose VLA-Forget, a hybrid unlearning framework that combines ratio-aware selective editing for perception and cross-modal specificity with layer-selective reasoning/action unlear...