Grounded and Faithful P&ID Reasoning: Constraining Vision-Language Models with Recovered Evidence Graphs
Graph-Grounded Harness (The authors') recovers an explicit graph of the drawing---its symbols, the process connections between them, and the tags that name them---and then requires 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 q...
P. Gadekar, Sagar Srinivas Sakhinana, Venkataramana Runkana
· 0 citations