Metraipolis › archive › dc_1328fed3d55f
MAE I Trust Myself? Self-Evaluating VLA Action Generation with Markov Attention Entropy — arXiv:2608.16697
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.16697
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T17:47:23.546Z
- SHA-256 of the archived text
3b049de2b5e877ae6eaa5a77bff71a2611c451fae2fa148492481ddb358dded3- Stable URL of this page
https://metraipolis.com/documento/dc_1328fed3d55f
How to cite
MAE I Trust Myself? Self-Evaluating VLA Action Generation with Markov Attention Entropy — arXiv:2608.16697 — arXiv (en), https://arxiv.org/abs/2608.16697 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1328fed3d55f (Metraipolis archive dc_1328fed3d55f, sha256 3b049de2b5e877ae6eaa5a77bff71a2611c451fae2fa148492481ddb358dded3)
BibTeX
@misc{dc1328fed3d55f,
title = {MAE I Trust Myself? Self-Evaluating VLA Action Generation with Markov Attention Entropy — arXiv:2608.16697},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.16697},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1328fed3d55f; sha256 3b049de2b5e877ae6eaa5a77bff71a2611c451fae2fa148492481ddb358dded3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.16697. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
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. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.16697 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-18T16:29:13.888Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).