aiqbench analysis
This paper introduces RA-VLA, a retrieval-augmented framework designed to overcome the adaptation bottlenecks inherent in standard Vision-Language-Action models when facing novel robotic tasks. By integrating behavior-aligned context retrieval with a grounded execution pipeline, the approach enables training-free task adaptation while maintaining inference efficiency on benchmarks like LIBERO and real-world UR5e environments. The work addresses critical limitations in current VLA architectures regarding behavioral inertia and superficial retrieval mechanisms, offering a scalable solution for robust general-purpose robotic manipulation.
RA-VLA: Retrieval-Augmented VLA for Test-Time Adaptation
arXiv:2608.25585v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models provide a versatile foundation for general robotic manipulation, yet they exhibit significant brittleness when confronted with novel task distributions. While In-Context Imitation Learning (ICIL) offers a training-free alternative, existing frameworks suffer from an adaptation bottleneck that hinders the effective translation of expert context to executable actions. This failure originates from superficial retrieval mechanisms and an inherent behavioral inertia that anchors the policy to its pre-trained priors. To address these limitations, we present RA-VLA, a retrieval-augmented VLA framework that integrates behavior-aligned context retrieval with a grounded execution pipeline. By enforcing faithful adherence to functional cues within a scalable architecture, RA-VLA facilitates seamless task adaptation while preserving inference efficiency. Our empirical evaluations across the LIBERO benchmark and a real-world UR5e environment demonstrate that RA-VLA achieves superior success rates and computational efficiency, establishing a robust framework for training-free robotic adaptation.
Abstract:Vision-Language-Action (VLA) models provide a versatile foundation for general robotic manipulation, yet they exhibit significant brittleness when confronted with novel task distributions. While In-Context Imitation Learning (ICIL) offers a training-free alternative, existing frameworks suffer from an adaptation bottleneck that hinders the effective translation of expert context to executable actions. This failure originates from superficial retrieval mechanisms and an inherent behavioral inertia that anchors the policy to its pre-trained priors. To address these limitations, we present RA-VLA, a retrieval-augmented VLA framework that integrates behavior-aligned context retrieval with a grounded execution pipeline. By enforcing faithful adherence to functional cues within a scalable architecture, RA-VLA facilitates seamless task adaptation while preserving inference efficiency. Our empirical evaluations across the LIBERO benchmark and a real-world UR5e environment demonstrate that RA-VLA achieves superior success rates and computational efficiency, establishing a robust framework for training-free robotic adaptation.
ICML 2026. Contact: this http URL@postech.this http URL
Robotics (cs.RO)
arXiv:2608.25585 [cs.RO]
(or arXiv:2608.25585v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.25585 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 · 241 words
Related coverage
Retrieve-then-Adapt: Retrieval-Augmented Test-Time Adaptation for Sequential Recommendation
arXiv · Aug 17, 2026
EASE-TTT: Evidence-Aligned Selective Test-Time Training for Long-Context Question Answering
arXiv · Jun 8, 2026
AgentOps: Operationalize agentic AI at scale with Amazon Bedrock AgentCore
AWS · Jun 1, 2026
When Does Retrieval Help? A Study of In-Context Adaptation in Vision-Language-Action Models
arXiv · Oct 6, 2026