N Neurarch Architectures Models Checks Data Docs Open the app

Resources / Datasets

Neurarch structural check catalogue

The 50 structural checks Neurarch runs on a model graph, as data: id, severity, category, the condition that triggers it, why it costs something, and the fix. 43 of them run on one graph, in the editor as a design is built, in CI through the GitHub Action, and over the wire at POST /api/v1/check; the other 7 compare a change with its base over POST /api/v1/plan and in the pull request comment. The catalogue is the vocabulary all of those surfaces grade in.

50 checks CC-BY-4.0 Free to use Read the catalogue On Hugging Face

Get it

The catalogue as JSONneurarch.com/r/index.json
application/json
Hugging Face mirror (dataset viewer, load_dataset)huggingface.co/datasets/neurarch-ai/neurarch-structural-checks
text/html
 
curl -s https://neurarch.com/r/index.json | jq '.checks[] | select(.severity=="blocker") | .id'

What is in a row

check id severity category trigger condition provenance

What this dataset is not

Severity is our editorial judgment, not a measured error rate. 3 of the 50 carry a non-null provenance field naming the study that grounds them; the rest follow from the shape algebra or from convention, and the entry says which. Reading the whole catalogue as measured would overstate it by 47 checks.

Licence and citation

Released under Creative Commons Attribution 4.0. Cite it as:

Neurarch. Neurarch structural check catalogue. https://neurarch.com/d/checks.html

The rest of the set

Neurarch architecture corpus
48 neural network architectures kept as typed graphs rather than diagrams.
48 architectures · CC0-1.0
Verifier grounding study (264 graphs)
Clean reference architectures plus systematically corrupted variants (broken attention head divisibility, linear width mismatches, severed connections), each built as a real PyTorch model and run on a GPU.
264 graphs · MIT
Arch-Bench arena results
Frontier language models scored on design-from-spec tasks by a deterministic verifier rather than a human or an LLM judge.
18 model-split results · CC-BY-4.0
Arch-Bench task set
The task definitions behind the benchmark: design-from-spec and repair-and-extend instances for agents that build neural network architectures.
12 curated tasks, 8 fixtures · MIT
arch-design-sft: verified architecture-design SFT data
Supervised fine-tuning data for neural architecture design treated as structured graph editing.
3,010 verified examples · MIT
Structure, verdict and trained outcome triples
The corpus that pairs what a design looks like with what it did.
80 trained graphs · MIT
Design selection replay: two trained campaigns and every selector tried
The records needed to score a design selector offline, in the shape of xRouteBench: every candidate design was trained once and the outcome kept, so a rule that picks which unexecuted design gets the GPU is evaluated by replay against the same rows as every previous rule, with no GPU and no model call.
39 trained designs · CC-BY-4.0
Who wrote the design: verifier refusal rate by author
The same pre-flight verifier applied to model designs written by three kinds of author, with the refusal rate per author kept apart.
7 rows · CC-BY-4.0
Verified architecture-design reasoning traces (Claude)
Spec to reasoning to design triples where the design is re-graded by the same deterministic verifier the benchmark uses, and only passing traces are kept.
306 verified traces · MIT
Verified architecture-design reasoning traces (Grok)
The same verified spec to reasoning to design triples as the Claude split, rejection-sampled from a different frontier model, so the two can be pooled for volume or held apart to see how much of the reasoning style is model-specific.
376 verified traces · MIT