Adaptive Capacity Allocation for Vision Language Action Fine-tuning

TL;DR

This paper introduces LoRA-SP, a rank-adaptive fine-tuning method for vision language action models that dynamically allocates capacity, leading to improved generalization and efficiency in robotic manipulation tasks.

Contribution

LoRA-SP adaptively allocates capacity during fine-tuning using an energy-based selection, outperforming fixed-rank methods in robotic vision language models.

Findings

LoRA-SP matches or exceeds full fine-tuning performance with fewer parameters.

It improves multi-task success rates by up to 31.6%.

The method is robust to rank choice and reduces cross-task interference.

Abstract

Vision language action models (VLAs) are increasingly used for Physical AI, but deploying a pre-trained VLA model to unseen environments, embodiments, or tasks still requires adaptation. Parameter-efficient fine-tuning (PEFT), especially LoRA, is common for VLA policies, yet the exposed capacity knob, the rank, does not transfer uniformly: robotics transfer exhibits a higher and task-varying intrinsic rank than language fine-tuning. Small ranks suffice for LLMs (e.g., $r \in \{4, 8\}$), while spectral analyses indicate VLAs may require much larger ranks (e.g., $r \approx 128$) or near-full rank, a mismatch that worsens in multi-task settings. We present LoRA-SP (Select-Prune), a rank-adaptive fine-tuning method that replaces fixed-rank updates with input- and layer-wise capacity. LoRA-SP uses an SVD-style parameterization with a small router whose nonnegative scores act as singular values over a shared vector bank. The active set is chosen by an energy target on the cumulative squared scores $E(k) \ge \eta$, providing a direct link to approximation error via our spectral analysis. During training, $\eta$ concentrates energy on a few directions and teaches the router to rely on fewer vectors while preserving accuracy. This yields compact adapters that reduce cross-task interference and improve generalization. On four real-robot manipulation tasks collected on an unseen AgileX PiPER arm, across two VLA backbones ($\pi_0$ and SmolVLA), LoRA-SP matches or exceeds full fine-tuning with far fewer trainable parameters, and improves multi-task success by up to 31.6% over standard LoRA while remaining robust to rank choice.