aiqbench analysis
ForgetMimic introduces a novel approach to "motion unlearning" for reinforcement learning policies in humanoid robots, directly addressing safety and privacy needs like GDPR compliance by removing specific behaviors without retraining the entire model. The method successfully degrades performance on targeted motions (e.g., dance or fight) while preserving capabilities for all others, resolving key failure modes inherent in standard robot control training mechanisms. This advancement is critical for deploying autonomous humanoids in regulated environments where data provenance and the right to be forgotten are paramount.
ForgetMimic: Motion Unlearning for Reinforcement Learning Humanoid Control
arXiv:2609.28378v1 Announce Type: new Abstract: Humanoid control, leveraging human demonstrations, has achieved diverse, agile, and natural locomotion behaviors through reinforcement learning (RL). While this paradigm has yielded remarkable performance in physical humanoid control, how to eliminate specific motions from learned policies remains insufficiently explored. Addressing this issue is motivated by pressing safety and privacy concerns: the removal of malicious, poisoned, or suboptimal motions, as well as copyright-protected motions subject to the right to be forgotten under regulations such as the GDPR, is of critical importance. To this end, we propose {ForgetMimic}, the first motion-level unlearning method designed specifically for physical-world humanoid control. The core idea of ForgetMimic is as follows: given a policy $\pi_\theta$ trained on $N$ motions, our method degrades performance on a target subset of $K$ motions while preserving the effectiveness of the remaining $N-K$ motions. Furthermore, we identify and resolve two key training mechanisms in robot control that lead to unlearning failure. We conduct extensive experiments on the Unitree G1 and H2 humanoid robots across 12 motions, including Dance, Fight, Flip, and others. Experimental results demonstrate that ForgetMimic effectively eliminates memory of designated motions while maintaining the normal operation of all other motions.
Abstract:Humanoid control, leveraging human demonstrations, has achieved diverse, agile, and natural locomotion behaviors through reinforcement learning (RL). While this paradigm has yielded remarkable performance in physical humanoid control, how to eliminate specific motions from learned policies remains insufficiently explored. Addressing this issue is motivated by pressing safety and privacy concerns: the removal of malicious, poisoned, or suboptimal motions, as well as copyright-protected motions subject to the right to be forgotten under regulations such as the GDPR, is of critical importance. To this end, we propose {ForgetMimic}, the first motion-level unlearning method designed specifically for physical-world humanoid control. The core idea of ForgetMimic is as follows: given a policy $\pi_\theta$ trained on $N$ motions, our method degrades performance on a target subset of $K$ motions while preserving the effectiveness of the remaining $N-K$ motions. Furthermore, we identify and resolve two key training mechanisms in robot control that lead to unlearning failure. We conduct extensive experiments on the Unitree G1 and H2 humanoid robots across 12 motions, including Dance, Fight, Flip, and others. Experimental results demonstrate that ForgetMimic effectively eliminates memory of designated motions while maintaining the normal operation of all other motions.
Robotics (cs.RO); Cryptography and Security (cs.CR); Machine Learning (cs.LG)
arXiv:2609.28378 [cs.RO]
(or arXiv:2609.28378v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.28378 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 291 words
Related coverage
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
The State of Simulation for Physical AI: An Overview
Hugging Face · Jul 21, 2026
REFINE-DP: Diffusion Policy Fine-tuning for Humanoid Loco-manipulation via Reinforcement Learning
arXiv · Jul 31, 2026
PLAT: Sparse Timed Keyframe Motion Tracking for Humanoid Control via Privileged Latent Transition Learning
arXiv · Sep 24, 2026