Metraipolis › archive › dc_bb9a99a56e7d
IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies — arXiv:2609.15005
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.15005
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T22:53:29.940Z
- SHA-256 of the archived text
974d51f9f930d14e33b289b99cb551b4a446e5cccb0f2363c1176be8a9c6afa9- Stable URL of this page
https://metraipolis.com/documento/dc_bb9a99a56e7d
How to cite
IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies — arXiv:2609.15005 — arXiv (en), https://arxiv.org/abs/2609.15005 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bb9a99a56e7d (Metraipolis archive dc_bb9a99a56e7d, sha256 974d51f9f930d14e33b289b99cb551b4a446e5cccb0f2363c1176be8a9c6afa9)
BibTeX
@misc{dcbb9a99a56e7d,
title = {IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies — arXiv:2609.15005},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.15005},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bb9a99a56e7d; sha256 974d51f9f930d14e33b289b99cb551b4a446e5cccb0f2363c1176be8a9c6afa9},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.15005. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Vision-Language-Action (VLA) policies perform robot manipulation tasks using multimodal inputs such as visual observations, proprioceptive states, and language instructions. However, it remains unclear at which execution stages each modality contributes to final task success and how input interventions propagate through subsequent states, observations, and actions. Existing attribution approaches primarily measure local sensitivity or temporally aggregated importance, limiting their ability to capture phase-dependent contributions and cross-phase dependencies. We propose Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies (IMPACT-VLA). IMPACT-VLA constructs behavioral phases from action transitions in a successful reference rollout, aligns them with policy query boundaries, and defines phase-modality blocks as attribution units. It then performs closed-loop counterfactual re-execution to quantify each block's contribution to final task success. We further analyze cross-phase non-additive interactions and trajectory propagation while distinguishing behavioral from functional recovery. Across 30 LIBERO robot manipulation tasks using OpenVLA-OFT, dominant-modality transitions occurred in 25 tasks (83.3%), and closed-loop attribution identified task-critical information more faithfully than Static Action Perturbation. Later-block marginal gains for negatively interacting pairs increased by approximately 3.3x under early-phase input replacement, while functional recovery could occur without behavioral recovery. These results reveal when multimodal inputs support task success and how their contributions become conditionally coupled during closed-loop execution. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-15 in cs.AI). Categorie arXiv: Robotics (cs.RO) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.15005 — 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 2026-10-15T22:29:03.769Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-15?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).