基于规则排序提示词的零样本多类别晶圆图分类：面向大型多模态模型

Zero-Shot Multi-Class Wafer Map Classification with Rule-Ordered Prompts for Large Multimodal Models

AI 导读

一种无需训练、面向冻结大型多模态模型的六类晶圆图缺陷分类框架，结合连通域高亮与形态、位置、排除感知提示词。在 Mixed-WM38（6,866 张图）上取得 92.9% macro 与 93.8% 总体准确率，在 WM811K（9,000 张图）上分别为 79.1% 和 81.3%。八种替代规则顺序的总体准确率为 90.1%-91.9%，低于所选顺序的 93.8%。

正文 · 原文

View PDF HTML (experimental)

Abstract:Wafer-map defect classification often requires labeled data and retraining when production conditions change. We present a training-free six-class framework that combines connected-component highlighting with morphology-, location-, and exclusion-aware prompts for a frozen large multimodal model. On Mixed-WM38 (6,866 maps), multi-region prompting achieves 92.9% macro and 93.8% overall accuracy. On WM811K (9,000 maps), it achieves 79.1% and 81.3%, respectively. Eight alternative rule orders achieve 90.1-91.9% overall accuracy, compared with 93.8% for the selected order.

Comments:	7 pages, 4 figures, 4 tables. Preprint version
Subjects:	Artificial Intelligence (cs.AI)
Cite as:	arXiv:2610.35080 [cs.AI]
	(or arXiv:2610.35080v1 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2610.35080 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Cao Tianjian [view email]
[v1] Mon, 26 Oct 2026 09:24:24 UTC (405 KB)

来源：arXiv cs.AI · arxiv.org