aiqbench analysis
PhaseLoRA introduces a novel parameter-efficient fine-tuning method that dynamically adjusts low-rank updates based on the temporal phases of robotic manipulation tasks, such as grasping or transport. By conditioning adaptation steps with descriptors like control tendency and event intensity, this approach significantly outperforms static LoRA baselines in continuous-action vision-language policies without freezing the model backbone entirely. This work highlights a critical advancement in making large-scale VLA models more adaptable to specific robot interaction regimes through lightweight architectural modifications.
PhaseLoRA: Control-Regime-Conditioned Low-Rank Adaptation for Continuous-Action Vision-Language-Action Policies
arXiv:2608.15285v1 Announce Type: new Abstract: Parameter-efficient fine-tuning (PEFT) is a natural way to adapt pretrained vision-language-action (VLA) policies, but most adapter designs apply temporally static updates throughout a control rollout, overlooking the phase-dependent nature of continuous-action manipulation. Such policies traverse distinct regimes, including approach, contact transition, grasping, transport, and placement, each requiring different adaptation behaviors. We propose \textbf{PhaseLoRA}, a lightweight LoRA parameterization that conditions adaptation at each action-chunk prediction step using two weakly supervised descriptors: fine-control tendency and event/boundary intensity. PhaseLoRA modulates the LoRA left factor in the action expert, allowing the effective low-rank update direction to vary over time while keeping the backbone largely frozen. On LIBERO, PhaseLoRA improves average success rate by 12.2 points over a matched-parameter high-rank LoRA baseline and outperforms stronger LoRA variants. Ablations show that random temporal modulation and scalar gating do not reproduce the performance of the full model, while update-direction analyses reveal structured temporal variation associated with the predicted control descriptors. These results establish within-trajectory conditioning as an effective lightweight PEFT axis for continuous-action VLA policies.
Abstract:Parameter-efficient fine-tuning (PEFT) is a natural way to adapt pretrained vision-language-action (VLA) policies, but most adapter designs apply temporally static updates throughout a control rollout, overlooking the phase-dependent nature of continuous-action manipulation. Such policies traverse distinct regimes, including approach, contact transition, grasping, transport, and placement, each requiring different adaptation behaviors. We propose \textbf{PhaseLoRA}, a lightweight LoRA parameterization that conditions adaptation at each action-chunk prediction step using two weakly supervised descriptors: fine-control tendency and event/boundary intensity. PhaseLoRA modulates the LoRA left factor in the action expert, allowing the effective low-rank update direction to vary over time while keeping the backbone largely frozen. On LIBERO, PhaseLoRA improves average success rate by 12.2 points over a matched-parameter high-rank LoRA baseline and outperforms stronger LoRA variants. Ablations show that random temporal modulation and scalar gating do not reproduce the performance of the full model, while update-direction analyses reveal structured temporal variation associated with the predicted control descriptors. These results establish within-trajectory conditioning as an effective lightweight PEFT axis for continuous-action VLA policies.
Robotics (cs.RO); Artificial Intelligence (cs.AI)
arXiv:2608.15285 [cs.RO]
(or arXiv:2608.15285v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.15285 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 · 264 words
Related coverage
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · Sep 4, 2026
Koopman Theory and Metaethics
LessWrong · Sep 19, 2026
How Would We Know? Reflections on trying to make AI go well amid deep uncertainty
LessWrong · Oct 5, 2026
The current bottleneck is political will, not research
AI Alignment Forum · Jul 11, 2026