WeaveLA: Event Driven Cross-Subtask Latent Memory Weaving for Repetitive Robot Manipulation

2026-06-15

AI Summary

Existing vision-language-action (VLA) policies exhibit fragility in multi-stage repetitive tasks due to the absence of a mechanism for cross-subtask information transfer. This work proposes an event-driven cross-subtask memory mechanism that, operating on a frozen VLA backbone, leverages subtask completion events as temporal units to compress historical states into implicit tokens via query-driven attention pooling. These tokens are then directly injected into the action generation pathway of the subsequent subtask, enabling lightweight and precise implicit memory transfer without disrupting the original short-horizon interface. Evaluated on the most challenging SwingXtimes (N=3) task in RoboMME, the approach boosts success rates from 0% to 47.8% while preserving performance on single-execution tasks.

Abstract

Vision-Language-Action (VLA) policies have achieved remarkable single-step manipulation, yet they remain brittle precisely where each stage depends on what was just completed. The core issue is structural: short-window VLAs lack an explicit channel for rouxting information across sub-task boundaries, and existing memory-augmented variants either write at every frame, retrieve from demonstration-time stages, or fire at sub-goal events without performing an explicit sub-task-to-sub-task hand-off into the action expert. We identify the sub-goal completion event as the natural temporal unit for cross-subtask memory hand-off, and present WeaveLA (Weave Latent memory for Vision-Language-Action policies), a cross-subtask memory interface that, on top of a frozen VLA backbone, compresses each completed segment into latent tokens via query-driven attention pooling and routes them directly into the action-generation path of the next sub-task. This event-triggered, action-side design preserves the base policy's short-window interface while adding a lightweight cross-subtask channel. Through stratified evaluation on RoboMME with a $π_{0.5}$ backbone, WeaveLA's gains land exactly where the channel is needed: on the hardest repetition slice (SwingXtimes, $N{=}3$), success rises from $0\%$ to $47.8\%$, while single-execution episodes remain unchanged. Per-episode paired analysis confirms the gains are confined to tasks whose causal structure requires cross-subtask information.

Problem

• Vision-Language-Action
• cross-subtask memory
• robot manipulation
• sub-goal completion
• repetitive tasks

Innovation

• event-driven memory
• cross-subtask reasoning
• latent memory weaving
• vision-language-action policies
• query-driven attention