# VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models

arXiv: https://arxiv.org/abs/2604.03956 · HTML: https://mlsanity.com/paper/2604.03956/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Generalist VLA models and benchmarks, one of 13 directions in Vision-language-action models.
- Builds on 25 of its 34 references that are in the map.
- First posted on arXiv 5 Apr 2026.

## Placed in

- Vision-language-action models / [Generalist VLA models and benchmarks](https://mlsanity.com/vla/generalist-vla-models-and-benchmarks/)
- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 25 of 25)

- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [OpenVLA: An Open-Source Vision-Language-Action Model](https://mlsanity.com/paper/2406.09246/) (2024)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [Gradient Surgery for Multi-Task Learning](https://mlsanity.com/paper/2001.06782/) (2020)
- [Machine Unlearning](https://mlsanity.com/paper/1912.03817/) (2019)
- [Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success](https://mlsanity.com/paper/2502.19645/) (2025)
- [Membership Inference Attacks From First Principles](https://mlsanity.com/paper/2112.03570/) (2021)
- [TOFU: A Task of Fictitious Unlearning for LLMs](https://mlsanity.com/paper/2401.06121/) (2024)
- [Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning](https://mlsanity.com/paper/2404.05868/) (2024)
- [VIMA: General Robot Manipulation with Multimodal Prompts](https://mlsanity.com/paper/2210.03094/) (2022)
- [Unified Concept Editing in Diffusion Models](https://mlsanity.com/paper/2308.14761/) (2023)
- [SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation](https://mlsanity.com/paper/2310.12508/) (2023)
- [Fast Machine Unlearning Without Retraining Through Selective Synaptic Dampening](https://mlsanity.com/paper/2308.07707/) (2023)
- [HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Model](https://mlsanity.com/paper/2503.10631/) (2025)
- [Catastrophic Failure of LLM Unlearning via Quantization](https://mlsanity.com/paper/2410.16454/) (2024)
- [SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning](https://mlsanity.com/paper/2509.05614/) (2025)
- [Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning rate](https://mlsanity.com/paper/2410.22086/) (2024)
- [Targeted Unlearning with Single Layer Unlearning Gradient](https://mlsanity.com/paper/2407.11867/) (2024)
- [LoTUS: Large-Scale Machine Unlearning with a Taste of Uncertainty](https://mlsanity.com/paper/2503.18314/) (2025)
- [Position: LLMs Must Use Functor-Based and RAG-Driven Bias Mitigation for Fairness](https://mlsanity.com/paper/2603.07368/) (2026)
- [CatRAG: Functor-Guided Structural Debiasing with Retrieval Augmentation for Fair LLMs](https://mlsanity.com/paper/2603.21524/) (2026)
- [G-Drift MIA: Membership Inference via Gradient-Induced Feature Drift in LLMs](https://mlsanity.com/paper/2604.00419/) (2026)
- [Improving Robustness of Vision-Language-Action Models by Restoring Corrupted Visual Inputs](https://mlsanity.com/paper/2602.01158/) (2026)
- [From Knowing to Doing Precisely: A General Self-Correction and Termination Framework for VLA models](https://mlsanity.com/paper/2602.01811/) (2026)
- [KORAL: Knowledge Graph Guided LLM Reasoning for SSD Operational Analysis](https://mlsanity.com/paper/2602.10246/) (2026)

## Built on by (most cited first, 10 of 10)

- [Vision-Language-Action Safety: Threats, Challenges, Evaluations, and Mechanisms](https://mlsanity.com/paper/2604.23775/) (2026)
- [PERSA: Reinforcement Learning for Professor-Style Personalized Feedback with LLMs](https://mlsanity.com/paper/2605.01123/) (2026)
- [SafeCommit: Certifying When Memory-Grounded Agents May Safely Act](https://mlsanity.com/paper/2608.04289/) (2026)
- [Listening with Attention: Entropy-Guided Explainability for Transformer-Based Audio Models](https://mlsanity.com/paper/2606.14647/) (2026)
- [Forgetting Only What Matters: Layer-Selective Unlearning toward Robust LLMs](https://mlsanity.com/paper/2609.10439/) (2026)
- [Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies](https://mlsanity.com/paper/2608.04692/) (2026)
- [Rethinking Demonstration Unlearning in Imitation Learning for Robotics](https://mlsanity.com/paper/2608.20784/) (2026)
- [Security of Foundation-Model-Powered Embodied Agents: Attack Surfaces, Attacks, Defenses, and Evaluation](https://mlsanity.com/paper/2608.16843/) (2026)
- [Security and Privacy in Large-Model-Driven Embodied Agents: Attacks, Defenses, and Future Directions](https://mlsanity.com/paper/2609.27847/) (2026)
- [REPORT-OPS: Hybrid Evidence-Guided Reporting with Human-Supervised Refinement for HPC](https://mlsanity.com/paper/2610.19222/) (2026)

## Abstract (arXiv)

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 unlearning for utility-preserving forgetting. VLA-Forget jointly optimizes three objectives: targeted forgetting, perceptual preservation, and reasoning retention, through staged updates over the visual encoder, projector, and upper action-generating transformer blocks. Across forget-set behavior probes and retain-task evaluations, VLA-Forget improves forgetting efficacy by 10%, preserves perceptual specificity by 22%, retains reasoning and task success by 9%, and reduces post-quantization recovery by 55% relative to strong unlearning baselines.