Retrieve in Time, Correct in Frequency

2026-08-05

AI Summary

Long-horizon vision-language-action (VLA) policies are prone to performance degradation due to error accumulation and visual ambiguities across task stages, leading to misaligned action progress. This work proposes a training-free, test-time correction framework that decouples action refinement into temporal alignment and frequency-domain adjustment. Specifically, it employs Progressive Memory Alignment (PMA) for unsupervised stage matching and injects clipped residuals into low-frequency motion channels to correct the outputs of a frozen policy. Relying solely on lightweight CPU computation, the method supports monotonic frontier-based incremental updates. Evaluated across four LIBERO suites, it improves average success rates from 86.4% to 88.4%, with a notable 7.0-percentage-point gain on LIBERO-Long (reaching 68.6%), while incurring a median latency of only 10.99 ms per correction.

Abstract

Frozen vision-language-action (VLA) policies generate temporally extended action chunks, but long-horizon manipulation remains vulnerable to accumulated execution error and visual aliasing across task stages. Successful rollouts provide useful corrective evidence, yet current frame retrieval can return progress-misaligned actions,while direct replay or time-domain fusion can overwrite the reactive structure of the policy proposal. We introduce Retrieve in Time, Correct in Frequency (RTCF), a training-free test-time correction framework that improves frozen VLA performance with low model-side overhead.RTCF separates which experience to retrieve from which part of its action to transfer. Progressive Memory Alignment (PMA) causally aligns the growing visual execution history with complete successful trajectories through incrementally updated monotonic frontiers, jointly identifying a relevant memory and the current aligned memory position without stage labels. From the aligned action chunk,RTCF transfers a coefficient-wise-clipped low-frequency residual on motion channels. Higher-frequency components and gripper decisions remain inherited from the frozen policy. Across four LIBERO suites and 2,000 episodes per condition, RTCF raises aggregate success from 86.4% to 88.4% and improves LIBERO-Long from 61.6% to 68.6%.These gains require no parameter updates, repeated VLA inference, or additional GPU resources: correction can be performed on the client CPU after a single policy invocation, and the median latencies sum to only 10.99 ms per action chunk

Problem

• execution error
• visual aliasing
• long-horizon manipulation
• frozen VLA policies
• action chunk alignment

Innovation

• Retrieve in Time, Correct in Frequency
• Progressive Memory Alignment
• frozen VLA policies
• low-frequency residual correction
• training-free test-time adaptation