扎实且可信的 P&ID 推理：利用恢复的证据图约束视觉语言模型
英文标题：Grounded and Faithful P&ID Reasoning: Constraining Vision-Language Models with Recovered Evidence Graphs
管道和仪表流程图（P&ID）是流程工业工厂的核心蓝图，设备隔离、维保与 HAZOP 等关键决策均高度依赖其管线拓扑连接，但现有视觉语言模型在此类复杂图纸上极易臆造或遗漏连接关系。为此，研究团队提出了一种新架构，先从工程图中提取包含图元符号、工艺连接及位号标签的显式拓扑证据图，并强制模型仅能通过 7 种只读算子查询该图来推导结论。在包含 3000 道图纸拓扑问答的全新基准 TopoPID-VQA 上，该框架将 Qwen3-VL 及 Gemma-4 等模型的精准匹配率从纯图像提示下的 36.7%–41.3% 大幅提升至 74.3%–76.0%。该研究为高风险工业工程图纸的可靠解析与自动化决策提供了兼具高可信度与可解释性的落地路径。
2.1 KB
461 B
612 B
380 B
9.1 KB
32.7 KB
22.8 KB
24.2 KB
34.7 KB
中文
[提交日期：2026 年 9 月 5 日]
标题：扎实且可信的 P&ID 推理：利用恢复的证据图约束视觉语言模型（Grounded and Faithful P&ID Reasoning: Constraining Vision-Language Models with Recovered Evidence Graphs）
查看由 Prathamesh Gadekar 等 3 位作者撰写的论文 PDF：《扎实且可信的 P&ID 推理：利用恢复的证据图约束视觉语言模型》
摘要：管道和仪表流程图（Piping and Instrumentation Diagrams，简称 P&ID）是流程工厂的权威图纸：设备隔离、检修维护以及 HAZOP 决策均取决于各组件之间的拓扑连接关系。视觉语言模型能够流畅地描述这些图纸，但它们往往会凭空臆造或遗漏工艺管线连接——而一条虚构或遗漏的连接就可能彻底颠覆隔离方案或可达性判断，因此工厂决策绝不能信赖未经图纸线条实际核对的流畅回答。为此，我们转而从工程图纸中还原出显式图（Graph）结构——包括其中的图元符号、符号间的工艺连接以及命名它们的位号标签——并要求模型仅能通过 7 种只读算子查询该拓扑图来给出回答，从而确保模型仅在能够引用支撑它的查询结果时才返回拓扑断言。在覆盖此类图纸的 3000 个全新拓扑问题基准评测集 TopoPID-VQA 上，我们提出的图扎实约束框架（Graph-Grounded Harness）将 Qwen3-VL-4B、Qwen3-VL-8B 和 Gemma-4-E4B 在纯图像提示下的精准匹配率（Exact Match）从 36.7%–41.3% 显著提升至 74.3%–76.0%。该方案是在并不完美的感知底座上实现的：在 Digitize-PID 数据集上，所恢复的拓扑图在精确工艺连接上的 F1 分数为 0.742，将符号和位号标签合并统计后 F1 分数达到 0.801。剩余的错误分布完全印证了这一感知差距——在恢复的图准确无误之处，证据扎实机制带来了丰厚回报；而在底层感知出错之处，拓扑问答仍会受到直接破坏。
提交历史
来自：Prathamesh Gadekar [ 查看邮箱 ]
[v1] 2026 年 9 月 5 日，星期六 05:02:34 UTC（3,794 KB）
English / 原文
[Submitted on 5 Sep 2026]
Title:Grounded and Faithful P&ID Reasoning: Constraining Vision-Language Models with Recovered Evidence Graphs
View a PDF of the paper titled Grounded and Faithful P&ID Reasoning: Constraining Vision-Language Models with Recovered Evidence Graphs, by Prathamesh Gadekar and 2 other authors
Abstract:Piping and Instrumentation Diagrams (P&IDs) are the authoritative maps of process plants: isolation, maintenance, and HAZOP decisions depend on what connects to what. Vision-language models describe these sheets fluently, yet they often invent or miss process connections---and an invented or missed link can reverse an isolation or reachability call, so a plant decision cannot trust a fluent answer that was never checked against the linework. We instead recover an explicit graph of the drawing---its symbols, the process connections between them, and the tags that name them---and then require the model to answer only by querying that graph through seven read-only operators, so a topology claim is returned only when it cites the query results that support it. On TopoPID-VQA, a new suite of 3000 topology questions over these sheets, Graph-Grounded Harness (Ours) raises exact match accuracy from 36.7--41.3% under image-only prompting to 74.3--76.0% for Qwen3-VL-4B, Qwen3-VL-8B, and Gemma-4-E4B. It does so on an imperfect substrate: on Digitize-PID dataset the recovered graph scores F1 0.742 on exact process connections, and 0.801 once symbols and tags are pooled in. The residual errors track that gap---grounding pays off where the recovered graph is right, and perception error still breaks topology questions where it is not.
Submission history
From: Prathamesh Gadekar [
]
[v1]
Sat, 5 Sep 2026 05:02:34 UTC (3,794 KB)