VLA Models Are More Generalizable Than You Think: Revisiting Physical and Spatial Modeling

2025/12/02 by Weiqi Li, Li, Weiqi, Qiang Zhang +6 · 1 citation
Computer Science · #Advanced Vision and Imaging #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Multimodal Machine Learning Applications #Robotics (cs.RO)

paper · pdf · doi:10.48550/arxiv.2512.02902

openalex publication_date 2025/12/02 · openalex created_date 2025/12/04 · openalex updated_date 2026/07/28

Abstract

Vision-language-action (VLA) models achieve strong in-distribution performance but degrade sharply under novel camera viewpoints and visual perturbations. We show that this brittleness primarily arises from misalignment in Spatial Modeling, rather than Physical Modeling. To address this, we propose a one-shot adaptation framework that recalibrates visual representations through lightweight, learnable updates. Our first method, Feature Token Modulation (FTM), applies a global affine transformation to visual tokens and improves Libero viewpoint accuracy from 48.5% to 87.1% with only 4K parameters. Building on this, Feature Linear Adaptation (FLA) introduces low-rank updates to the ViT encoder, achieving 90.8% success with 4.7M parameters -- matching LoRA-scale finetuning at far lower cost. Together, these results reveal substantial untapped robustness in pretrained VLA models and demonstrate that targeted, minimal visual adaptation is sufficient to restore viewpoint generalization.

Citations

• LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models
• LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorization
• GeoAware-VLA: Implicit Geometry Aware Vision-Language-Action Model
• Physical Autoregressive Model for Robotic Manipulation without Action Pretraining
• π_0.5: a Vision-Language-Action Model with Open-World Generalization
• Adapt3R: Adaptive 3D Scene Representation for Domain Transfer in Imitation Learning
• Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success
• VLABench: A Large-Scale Benchmark for Language-Conditioned Robotics Manipulation with Long-Horizon Reasoning Tasks
• 3D-VLA: A 3D Vision-Language-Action Generative World Model
• Vision-Language Foundation Models as Effective Robot Imitators
• GELLO: A General, Low-Cost, and Intuitive Teleoperation Framework for Robot Manipulators
• RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
• LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning
• PaLM-E: An Embodied Multimodal Language Model
• Conditional Prompt Learning for Vision-Language Models
• BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models
• LoRA: Low-Rank Adaptation of Large Language Models
• Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning
• Parameter-Efficient Transfer Learning for NLP
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Prefix-Tuning: Optimizing Continuous Prompts for Generation

Cited by

• From Fixed to Free Cameras: Calibration-Free View-Robust Vision-Language-Action Model

Related

• MAPS: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better Vision-Language-Action Generalization
• VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Model
• VLA-LPAF: Lightweight Perspective-Adaptive Fusion for Vision-Language-Action to Enable More Unconstrained Robotic Manipulation
• Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization
• Rethinking Fine-Tuning: Unlocking Hidden Capabilities in Vision-Language Models