Metraipolis › archive › dc_fece63ef1e87

The Layer Mystery of VLA: An Information-Theoretical Analysis of VLA Latent Interface — arXiv:2609.36118

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.36118
License (as archived)
CC0-1.0
Archived
2026-10-07T16:22:17.048Z
SHA-256 of the archived text
2ae7ebec93f06b00508a235778373119c0bdfadb7829004c6a6fb86c22cf4835
Stable URL of this page
https://metraipolis.com/documento/dc_fece63ef1e87

How to cite

The Layer Mystery of VLA: An Information-Theoretical Analysis of VLA Latent Interface — arXiv:2609.36118 — arXiv (en), https://arxiv.org/abs/2609.36118 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fece63ef1e87 (Metraipolis archive dc_fece63ef1e87, sha256 2ae7ebec93f06b00508a235778373119c0bdfadb7829004c6a6fb86c22cf4835)
BibTeX
@misc{dcfece63ef1e87,
  title = {The Layer Mystery of VLA: An Information-Theoretical Analysis of VLA Latent Interface — arXiv:2609.36118},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36118},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fece63ef1e87; sha256 2ae7ebec93f06b00508a235778373119c0bdfadb7829004c6a6fb86c22cf4835},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36118. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Vision-language-action (VLA) policies connect a pretrained vision-language backbone to an action head through a latent interface, but which backbone layers this interface should expose remains unclear. We study single-layer selection and multi-layer fusion for frozen backbones across three pretrained models and two manipulation benchmarks, LIBERO and CALVIN, with three policy-training seeds per configuration. Across three fusion mechanisms and three layer-subset strategies, 47 of 54 configurations underperform the best observed single-layer policy. Our stastical analysis further confirms that fusion's advantage is very limited. However, the best layer varies substantially across backbones and benchmarks, making layer selection consequential and exhaustive policy sweeps expensive. We further derive a reweighting equivalence between the proposed information-bottleneck objectives for action-conditioned InfoNCE and action prediction, motivating InfoNCE as a proxy for layer quality. Empirically, InfoNCE provides the most consistent positive association with policy success among four evaluated proxies. Selecting the layer with the highest InfoNCE score requires 9-33 times less GPU compute than exhaustive policy sweeps and reduces mean selection regret from 17.89 percentage points for deepest-layer selection to 3.71 points across six settings. Its mean regret is close to the 3.28-3.50 points achieved by fixed-layer heuristics optimized retrospectively using all six oracle sweeps, without requiring closed-loop evaluations during selection.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.36118 — 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 2026-10-07T15:54:53.468Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).