16‐1: Invited Paper: Visual Chain‐of‐Thought Reasoning for Display Industrial Defect Management Based on Vision‐Language Models

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 10.8% improvement over supervised fine‐tuned VLMs.