Metraipolis › archive › dc_dced9399819a

Readout is not Recovery: Dissociating Coordinate Emission from Visual-Corruption Repair in Vision-Language Models — arXiv:2609.32757

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.32757
License (as archived)
CC0-1.0
Archived
2026-10-07T20:56:02.806Z
SHA-256 of the archived text
abe9da0d0590da1ec415cf8508a5596a256bb31ba5454bf6a2b193fb6fe27bab
Stable URL of this page
https://metraipolis.com/documento/dc_dced9399819a

How to cite

Readout is not Recovery: Dissociating Coordinate Emission from Visual-Corruption Repair in Vision-Language Models — arXiv:2609.32757 — arXiv (en), https://arxiv.org/abs/2609.32757 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dced9399819a (Metraipolis archive dc_dced9399819a, sha256 abe9da0d0590da1ec415cf8508a5596a256bb31ba5454bf6a2b193fb6fe27bab)
BibTeX
@misc{dcdced9399819a,
  title = {Readout is not Recovery: Dissociating Coordinate Emission from Visual-Corruption Repair in Vision-Language Models — arXiv:2609.32757},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.32757},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dced9399819a; sha256 abe9da0d0590da1ec415cf8508a5596a256bb31ba5454bf6a2b193fb6fe27bab},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32757. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

VLM bounding-box localization is both language generation and spatial commitment. Parseable fields such as bbox_2d make localization easy to score, but dimensions that emit coordinate tokens need not repair localization after visual evidence is damaged. We study this readout/recovery separation in Qwen3-VL-4B-Instruct on single-object COCO grounding. We compare clean coordinate-token readout rankings with corruption-derived repair rankings, using object-mask endpoint replacement for recovery and clean-input flooring for depth localization. In Qwen3-VL, coordinate-token rankings are inert through layer 24, load-bearing from layers 32-35, and peak at layer 34; corruption-derived rankings harm layers 16-24 but become beneficial near layer 35/final. A Kimi-VL-A3B diagnostic shows a matching output-proximal transition despite a different box format. Object-mask recovery separates rank budgets: $k=250$ shows necessity, $k=500$ shows Top-$k$ restoration above random, and $k=d/2$ is largely capacity-driven. Partial-occlusion sweeps reveal that high-overlap coordinate-token sets can hurt at $k=1000$ and help mainly at half-width, while population corruption-derived sets provide no reliable fixed repair set. Edge-attribution patching shows coordinate-token paths are high precision but low recall for detection recovery, and RMSNorm quasi-layer controls do not close the endpoint-repair gap. Endpoint coordinate triage is therefore a useful circuit prior, but occlusion recovery requires a separate benchmark.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.32757 — 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 2026-10-07T20:29:00.963Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).