Metraipolis › archive › dc_0fd429fcd80b

DepthEvidence: Unifying Metric Depth Prediction and Geometric Reasoning in Multimodal Language Models — arXiv:2609.31103

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.31103
License (as archived)
CC0-1.0
Archived
2026-10-08T01:39:41.276Z
SHA-256 of the archived text
ee6c9c7818885c5293a26c1673312e89457d8ecaf24b82356b4219768fd4cad6
Stable URL of this page
https://metraipolis.com/documento/dc_0fd429fcd80b

How to cite

DepthEvidence: Unifying Metric Depth Prediction and Geometric Reasoning in Multimodal Language Models — arXiv:2609.31103 — arXiv (en), https://arxiv.org/abs/2609.31103 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0fd429fcd80b (Metraipolis archive dc_0fd429fcd80b, sha256 ee6c9c7818885c5293a26c1673312e89457d8ecaf24b82356b4219768fd4cad6)
BibTeX
@misc{dc0fd429fcd80b,
  title = {DepthEvidence: Unifying Metric Depth Prediction and Geometric Reasoning in Multimodal Language Models — arXiv:2609.31103},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.31103},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0fd429fcd80b; sha256 ee6c9c7818885c5293a26c1673312e89457d8ecaf24b82356b4219768fd4cad6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31103. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Spatial reasoning with metric constraints requires linking objects to geometric measurements and preserving their numerical content during language reasoning. We present DepthEvidence, a 4B model that uses its own dense metric predictions as object-grounded evidence for language generation. A camera-conditioned decoder predicts full-resolution metric depth using multi-scale visual features and high-resolution RGB refinement. A dense-to-language interface converts predicted depths and decoder features into object-aligned continuous geometry tokens anchored to object identifiers. Geometric supervision encourages metric information to remain recoverable before and after language-context interaction, while instruction tuning supports object measurement and compositional reasoning. We introduce a Depth-VQA benchmark evaluating object-depth queries, relative comparisons, and decisions combining spatial and numerical constraints. Across nine datasets, DepthEvidence achieves the highest average dense $\delta_1$ among evaluated methods, competitive with specialized estimators. It also leads the evaluated methods in instance-level metric depth estimation and overall accuracy on both relative and metric reasoning tracks, while broadly preserving general VQA performance and improving spatial understanding relative to the base model.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.31103 — 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 2026-10-08T01:12:14.531Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).