Metraipolis › archive › dc_ef79f89ff896

GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs — arXiv:2608.04510

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.04510
License (as archived): CC0-1.0
Archived: 2026-10-31T11:14:24.479Z
SHA-256 of the archived text: d1509bc6e83fc20895f65704434458732bd69de3d4118e4d0c59e17d8a192fc9
Stable URL of this page: https://metraipolis.com/documento/dc_ef79f89ff896

How to cite

GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs — arXiv:2608.04510 — arXiv (en), https://arxiv.org/abs/2608.04510 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ef79f89ff896 (Metraipolis archive dc_ef79f89ff896, sha256 d1509bc6e83fc20895f65704434458732bd69de3d4118e4d0c59e17d8a192fc9)

BibTeX

@misc{dcef79f89ff896,
  title = {GUARD: Grounding Uncertainty and Ablation-Based Risk Detection for Diffusion-Based VLAs — arXiv:2608.04510},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.04510},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ef79f89ff896; sha256 d1509bc6e83fc20895f65704434458732bd69de3d4118e4d0c59e17d8a192fc9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.04510. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Diffusion-based vision-language-action (VLA) policies can generate plausible actions even when their predictions are weakly grounded in the visual and language evidence defining the task. We introduce GUARD, a test-time failure detection method that measures this grounding without modifying the pretrained policy. GUARD estimates the influence of token-indexed entries in the final vision-language model key-value (KV) cache, constructs counterfactual caches by ablating salient KV entries, and compares their denoising responses with the original conditioning. Based on the comparison, we derive GUARD diagnostic stream including sensitivity, attention entropy, modality bias, and grounding efficiency, which are calibrated online and processed by a lightweight temporal classifier. We evaluate GUARD under task-held-out splits across five policy-benchmark settings, using Pi0, SmolVLA, and Alpamayo-1.5 on LIBERO, SimplerEnv, MetaWorld, and PhysicalAI-AV. GUARD achieves the best ROC-AUC on four of five unseen-task settings and ranks second on the remaining setting, improving the average unseen-task ROC-AUC by 5.73 percentage points over the strongest competing runtime monitor while remaining within 0.19 points of the best seen-task average. These results show that directly probing action-head dependence on multimodal evidence provides a transferable failure signal across policies, tasks, embodiments, and domains.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-06 in cs.AI).
Categorie arXiv: Robotics (cs.RO) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.04510 — 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 2026-10-31T11:02:13.523Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-06?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).