Metraipolis › archive › dc_74f09775fd99
Is In-Domain Training Enough for Fine-Grained Industrial Anomaly Understanding? — arXiv:2610.12310
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.12310
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T06:30:57.786Z
- SHA-256 of the archived text
e863581f421676909c457bead38711eb2d67aa2563b9a9c92fcd2d7e7df0b8c9- Stable URL of this page
https://metraipolis.com/documento/dc_74f09775fd99
How to cite
Is In-Domain Training Enough for Fine-Grained Industrial Anomaly Understanding? — arXiv:2610.12310 — arXiv (en), https://arxiv.org/abs/2610.12310 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_74f09775fd99 (Metraipolis archive dc_74f09775fd99, sha256 e863581f421676909c457bead38711eb2d67aa2563b9a9c92fcd2d7e7df0b8c9)
BibTeX
@misc{dc74f09775fd99,
title = {Is In-Domain Training Enough for Fine-Grained Industrial Anomaly Understanding? — arXiv:2610.12310},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.12310},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_74f09775fd99; sha256 e863581f421676909c457bead38711eb2d67aa2563b9a9c92fcd2d7e7df0b8c9},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12310. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A single multimodal large language model (MLLM) struggles to excel simultaneously at detection, localization, description, and reasoning in multimodal industrial anomaly understanding (MM-IAU). We show that in-domain training does not close this gap. On MMAD, a widely adopted MM-IAU benchmark, trained specialists reach at most 75.5% accuracy in defect localization, against 92.3% for human experts, and even detect anomalies less accurately than their untrained base model. Meanwhile, different MLLMs offer complementary strengths but share this weakness in fine-grained perception, so combining them alone cannot remove it. We therefore propose SiGMA, a spatially grounded multi-agent framework that divides labor between heterogeneous MLLM agents and a dedicated visual defect expert. A multimodal searcher supplies industrial knowledge and normal references, the defect expert turns query-reference comparison into calibrated anomaly evidence, and a label-free reliability controller weighs each source by task-wise competence and query-level evidence quality. SiGMA reaches 85.2% average accuracy on MMAD, 4.0% above the strongest trained specialist and Gemini-2.5-Pro and within 1.5% of human experts. Even with three agents of at most 9B parameters, it reaches 84.4%, and new MLLMs join without retraining. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2610.12310 — 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 2026-10-09T05:55:07.689Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).