ReCAPA: Hierarchical Predictive Correction to Mitigate Cascading Failures

TL;DR

ReCAPA introduces a hierarchical predictive correction framework for vision-language-action systems to reduce cascading failures and improve task execution accuracy in multimodal environments.

Contribution

It proposes a novel prediction and contrast-based alignment architecture with new metrics for error propagation and recovery in long-horizon tasks.

Findings

ReCAPA outperforms baseline models on VisualAgentBench, MineDojo, and AI2-THOR.

The framework effectively mitigates error propagation during multi-step task execution.

Experiments demonstrate improved alignment and task success rates.

Abstract

Vision-Language-Action systems follow instructions to execute multi-step tasks in multimodal environments. Recent VLA approaches typically rely on post-hoc correction mechanisms or operate under fixed task decompositions and alignment schemes. However, once an intermediate step is mis-specified, local errors propagate through subsequent steps and eventually accumulate into cascading failures. To mitigate this compounding effect, we propose Predictive Alignment and Planning Architecture, a framework that uses prediction and contrast to adjust deviations across three levels: actions, subgoals, and trajectories. Semantic alignment is enforced at all levels using a Sinkhorn-based module and a Score-field module. The predictive correction and alignment jointly update the action generator during training, enabling it to adjust fine-grained steps to remain aligned with the overall intent. We further introduce two new metrics to quantify error propagation and recovery processes in tasks, capturing how mistakes spread and fade over long-horizon execution. Experiments show that ReCAPA achieves competitive results on embodied agent benchmarks such as VisualAgentBench, MineDojo, and AI2-THOR, outperforming strong proprietary and open-source Large Language Model baselines.