ChainVLA: Chaining Vision-Language-Action Queries through a Unified Execution State for Long-Horizon Manipulation

2026-08-03

AI Summary

Existing vision-language-action (VLA) policies often suffer from fragmented planning in long-horizon tasks due to the lack of unified modeling of action continuity and task progress memory. This work proposes ChainVLA, which uniquely integrates both aspects into a unified revisable execution state: it constructs task progress context through recursive working memory and sparse event memory, while preserving unexecuted portions of prior actions as a “motion tail” to maintain action intent. The system dynamically updates this state with incoming observations and regenerates actions accordingly. Built upon a 1.2B-parameter VLA architecture, ChainVLA achieves an average success rate of 62.8% on RMBench and 98.8% across four LIBERO suites. Ablation studies confirm the critical role of each component, as removing the motion tail or progress context reduces RMBench performance drastically to 11.2% and 3.0%, respectively.

Abstract

Humans perform long-horizon manipulation by retaining knowledge of what earlier actions have established while continuously adapting the motion underway. By contrast, action-chunked vision-language-action (VLA) policies repeatedly replan from the current input at each query. Existing methods preserve either long-term task evidence through memory or short-term motion through action reuse and ensembling, leaving the cross-query handoff incomplete. We introduce ChainVLA, a 1.2B-parameter VLA policy that chains successive queries through a joint and revisable execution state. Progress Context combines a recurrent Working State with sparse event memory to carry observation-derived task progress, while Motion Tail feeds the preceding prediction's unexecuted continuation into state construction and action generation. Together, the two components condition a decoder that regenerates each action horizon under the latest observation, allowing the carried state to guide the next prediction without fixing it. ChainVLA reaches 62.8% average success on RMBench and 98.8% across four LIBERO suites, while removing Motion Tail or Progress Context reduces RMBench success to 11.2% and 3.0%, respectively. These asymmetric ablations are consistent with motion continuity helping preserve the observation stream from which task progress is inferred.

Problem

• long-horizon manipulation
• vision-language-action
• execution state
• action chaining
• task progress

Innovation

• ChainVLA
• execution state
• long-horizon manipulation
• vision-language-action
• motion continuity