AI P&ID tag-register checks
Check whether AI preserves equipment and instrument tags, revision states and explicit uncertainty before using extracted P&ID information in an asset register.
Synthetic evaluation input
tag,type,revision_state,note P-101,pump,current,first entry P-101,pump,current,duplicate entry PT-101,pressure transmitter,current,not P-101 P-102,pump,removed,removed in revision C L-10,line,current,destination not stated
Original Inzonex training example, version 1.0. No customer document or real equipment data. No model has been scored on this page.
Expected fields
| Field | Reference answer |
|---|---|
| unique_current_tags | ["P-101", "PT-101", "L-10"] |
| removed_tags | ["P-102"] |
| duplicate_tag | "P-101" |
| line_destination | null |
Check a structured response
The comparison runs in your browser against this answer key. It checks exact fields and values, not the accuracy of an explanation or a model's general ability.
Recommended workflow
Separate tag recognition from connectivity interpretation. Reconcile extracted tags against a reviewed register before attempting drawing-level relationships.
Reading a tag and understanding a process connection are different tasks. A tag register can be correct while an inferred connection is wrong. Keep these evaluations separate so a successful text extraction does not become evidence that the model understands a drawing.
The packet below is a synthetic text register, not a P&ID image. It tests duplicate handling, one-character tag differences and revision exclusions. A drawing benchmark also needs actual drawing crops, a reviewed legend and independently checked connections.
Treat removed and existing tags as different states. Retaining a removed component in the current register is a practical failure even when every character was recognised correctly. Unknown line destinations must stay unknown.
Failure checks
- Preserve PT-101 and P-101 as different tags.
- Deduplicate P-101 without losing the removed-tag record.
- Do not infer where L-10 connects.
What this exercise does not prove
This is register QA only. It does not benchmark symbol recognition, connectivity, process safety or drawing completeness.
For an actual evaluation, keep a separate held-out set, record tool/model version and settings, and log raw outputs, corrections, elapsed time and actual charges. Do not compare tools tested on different inputs as though they ran the same benchmark.
What is checked in AI P&ID tag-register checks?
Check whether AI preserves equipment and instrument tags, revision states and explicit uncertainty before using extracted P&ID information in an asset register.
What does AI P&ID tag-register checks not prove?
This is register QA only. It does not benchmark symbol recognition, connectivity, process safety or drawing completeness.
Related tasks
Match this workflow to your data requirements
Methodology and reuse
These packets and answer keys are original Inzonex educational material, licensed CC BY 4.0. Attribute Inzonex and link to this task page when reusing the packet. The licence does not cover third-party material linked from this site.
NIST AI 600-1: background on generative AI evaluation and risk. This exercise is not NIST-certified.