84‐3: Visual Chain‐of‐Thought Reasoning for Display Industrial Defect Management Based on Vision–Language Models
SID Symposium Digest of Technical Papers · 2026-06 · 57 (1) · 1167-1170
DOI: 10.1002/sdtp.70001_315 ↗Kurz gefasst
Fachartikel in der Zeitschrift SID Symposium Digest of Technical Papers, online erschienen am 11.09.2026. Verfasst von Haiyang Guo und 6 weiteren Autor:innen. Die Arbeit nennt 13 Literaturangaben.
Abstract (Originalsprache)
In this study, we propose an end‐to‐end intelligent defect management based on a vision–language model (VLM); this framework performs defect type detection, root cause analysis, and repair strategy generation. The main contributions are as follows: (1) construction of an instruction‐guided image–text pair dataset containing chain‐of‐thought reasoning information; (2) supervised fine‐tuning of VLMs with generated data; (3) integration of visual information as reward feedback during the reinforcement learning training of VLMs. Experimental results demonstrate that the method achieves an accuracy of 95.69% on the defect repair dataset, representing a 5.68% improvement over supervised fine‐tuned VLMs.
Zitieren
APA
Guo, H., Wang, T., Song, L., Yang, Y., Yin, S., Lv, B., & Jiang, X. (2026). 84‐3: Visual Chain‐of‐Thought Reasoning for Display Industrial Defect Management Based on Vision–Language Models. SID Symposium Digest of Technical Papers, 57(1), 1167-1170. https://doi.org/10.1002/sdtp.70001_315
Vancouver
Guo H, Wang T, Song L, Yang Y, Yin S, Lv B, et al. 84‐3: Visual Chain‐of‐Thought Reasoning for Display Industrial Defect Management Based on Vision–Language Models. SID Symposium Digest of Technical Papers. 2026;57(1):1167-1170. doi:10.1002/sdtp.70001_315
BibTeX
@article{Guo2026,
author = {Guo, Haiyang and Wang, Tingting and Song, Lei and Yang, Ye and Yin, Shujuan and Lv, Bin and Jiang, Xingqun},
title = {84‐3: Visual Chain‐of‐Thought Reasoning for Display Industrial Defect Management Based on Vision–Language Models},
journal = {SID Symposium Digest of Technical Papers},
year = {2026},
volume = {57},
number = {1},
pages = {1167--1170},
doi = {10.1002/sdtp.70001_315}
}
EUVIMED