arch-design-sft: verified architecture-design SFT data

Supervised fine-tuning data for neural architecture design treated as structured graph editing. Each row pairs a natural-language design spec and a serialized starting graph with a reference action plan, and every row is re-graded by a deterministic verifier before it is written: structural blockers, parameter budgets and bands, required layer families. Ten task families across six design-from-spec and four edit-in-place, where wholesale rebuilds are forbidden so the data teaches surgical edits rather than generation. Tasks are procedurally minted from a seed, so an evaluation split on a different seed has never existed anywhere.

• Size: 3,010 verified examples
• Licence: MIT License
• Canonical page: https://neurarch.com/d/arch-design-sft.html
• Longer write-up: https://github.com/neurarch-ai/neurarch-arch-bench/blob/main/training/DATASET_CARD.md

Fields

• task family
• seed
• design spec
• starting graph
• reference action plan
• verifier grade

What this dataset is not

The targets are machine-checked, not human-preferred. Every assistant turn provably satisfies its task constraints, and nothing in the pipeline claims it is the design an experienced engineer would have chosen. Procedural generation is what makes the split contamination-free and is also what makes it narrower than real design work.

Verify a design of your own

These rows describe neural network graphs that were checked by a deterministic verifier, and the same verifier is callable:

curl -X POST https://www.neurarch.com/api/v1/check \
  -H "Authorization: Bearer $NEURARCH_API_KEY" \
  -H "Content-Type: application/json" \
  -d @graph.json

Cite

Neurarch. arch-design-sft: verified architecture-design SFT data. https://neurarch.com/d/arch-design-sft.html