Construction RFIs paired with design team answers, evaluation pack

Summary

A curated evaluation pack of construction RFIs paired with the architect's or engineer's final answer, drawn from the project records of 21 US general and specialty contractors. Each task gives the question, the drawing and spec references it cites and the relevant spec text, and the reference answer is the design team's actual response, graded for completeness by a construction reviewer. Projects, people and firms are replaced with placeholders, and RFIs that depend on drawings the text cannot describe are removed.

Collected to order. We collect it from US companies for your order. Scope, volume and timing are confirmed in writing before you sign.

Pricing on request

Request scope and pricing

At a glance

  • Planned scope. Volume, sources and timing are confirmed in writing before you sign.
Records
18Kevaluation tasks
Size
220 MB
Tokens, estimated
37M
Years
2015 to 2025
Firms
21
Owner profile
21 US general and specialty contractors, 12 to 240 staff each, commercial, education and multifamily work
Industry
Construction
Data type
Evaluation and task packs
Headcount
12 to 240 staff per firm
Region
United States
Languages
English
Formats
JSONL
Updated

Sources

SystemKindRecords
Procore RFIsProjects and wiki61K RFIs screened, 18.4K kept
Shared drives (project manuals)Shared files9.8K spec sections

Good for

  • evaluating whether a model can answer an RFI from the spec text alone
  • measuring invented code and spec citations in construction answers
  • benchmarking retrieval over long project manuals
  • grading agent-drafted RFI responses against the design team's real answer

Fields

FieldTypeDescription
task_idstringUnique task id
disciplineenumarchitectural, structural, mechanical, electrical, plumbing, civil or fire_protection
questionstringRFI question as submitted
referencesarray<string>Drawing sheets and spec paragraphs the RFI cites
contextstringText of the cited spec section
reference_answerstringThe design team's final response
answer_typeenumclarification, substitution_accepted, substitution_rejected, design_change or refer_to_documents
completeness_gradeintReviewer grade of the reference answer, 1 to 5
splitenumdev or test

Example record

{
  "task_id": "RFI-EV-04417",
  "discipline": "plumbing",
  "question": "Spec 22 13 16 calls for cast iron for all sanitary piping above grade. Can PVC Schedule 40 DWV be used above grade in the 2-story classroom wing, where the ceiling space is not a return air plenum?",
  "references": [
    "P-201",
    "M-101 note 4",
    "22 13 16 2.3"
  ],
  "context": "2.3 ABOVE GRADE SANITARY PIPING. A. Hub-and-spigot or hubless cast iron soil pipe and fittings. B. Plastic piping is not permitted in return air plenums or rated assemblies.",
  "reference_answer": "PVC Schedule 40 DWV is acceptable above grade in the classroom wing only, where return air is ducted. Cast iron remains required in the gym and corridor ceilings, which are return plenums, and at all rated wall penetrations.",
  "answer_type": "substitution_accepted",
  "completeness_grade": 5,
  "split": "test"
}
One record, de-identified. Placeholders such as [PERSON_12] and [COMPANY_4] stay the same across the whole dataset, so threads still read correctly.

Licenses available

  • Non-exclusive

    Licensed to more than one buyer.

  • First look

    One buyer gets 2 to 4 weeks to make an offer before anyone else sees the dataset.

What ships with it

  • Datasheet: what it is, what it is good for, verified size, sources, years, format and limits.
  • Provenance file: the owner shown by code, staff notice and consent status, exclusions, mapped to the EU AI Act Article 53 training-data summary and California AB 2013.
  • Manifest with a hash for every file.
  • QC report and de-identification report.
  • Your own copy, with unique canary records.

How it is de-identified

Default exclusions removed, files flattened to text, identifiers found in three layers and replaced with consistent placeholders, and the result measured against a 99% bar for each identifier type in each source.

The full method

Request scope and pricing

We confirm scope, volume and timing in writing before you sign. Samples go out under evaluation terms after a short buyer check, once the first records are in.

Dataset: NK-25196

Organization type
What would you like?
Intended use
License interest

We use these details to reply to you and handle your request. See our Privacy Notice.