FabriMAE I Trust Myself? Self-Evaluating VLA Action Generation with Markov Attention Entropy
2026-08-17 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors study models that combine vision, language, and actions to perform tasks, called VLAs. They focus on helping these models check how reliable their own actions are without needing outside help. They find that looking at the models' internal attention patterns, specifically something called entropy, can tell if a task was done well or not. Using this idea, they create a new method called MAE that uses these signals to score action reliability. They tested MAE on a big new benchmark and showed it works better than previous methods at recognizing when actions might fail.
Vision-Language-Action modelsSelf-evaluationEntropyAttention mechanismMarkov ChainReliability scoreBenchmarkAction generationUncertainty estimationTest-time selection
Authors
Aniri, Chen Yilin, Jinhe Bi, Junfei Guo, Donglai Ran, Xu Bian, Zengjie Jin, Yujun Wang, Yijun Tian, Volker Tresp, Fei Shen, Tat-Seng Chua, Yunpu Ma
Abstract
Vision-Language-Action models (VLAs) integrate visual perception, language instruction, and action generation into end-to-end policies across heterogeneous architectures. However, enabling VLAs to self-evaluate their action generation reliability without external supervision remains a major challenge. Existing methods either rely on expert annotations or estimate uncertainty only from output statistics, largely ignoring internal signals. In this work, we observe that internal visual modality entropy exhibits consistent distinctions between successful and failed tasks across heterogeneous VLAs. Although VLAs' architectures differ in their action generation, we show that they share a common latent action generation abstraction evolving under visual perception, language instruction, and state input, which we formulate as a Conditional Generative Markov Chain. Based on this formulation, we propose MAE (Markov Attention Entropy), a self-evaluation framework that directly converts internal attention signals into architecture-aware reliability scores, and introduce LIBERO-Reflect, a 4,000-episode benchmark combining 2,000 standard episodes and 2,000 challenging episodes across four subsets. Extensive experiments across heterogeneous VLA architectures and diverse scenarios show that MAE consistently outperforms state-of-the-art baselines on AUPR, AUROC, and FPR@95. We further instantiate FabriMAE for verifier-free test-time action selection, showing that MAE-guided multiple sampling improves PI-family robustness on LIBERO-Plus with small observed runtime overhead.