Cross-check a material prediction against literature and historical data.
by Aigarth Validate
Cross-check a material prediction against literature and historical data. Returns a confidence score, the literature citations, and the historical mean ± σ. Catches predictions that are numerically correct but physically wrong.
Deploy is a Phase 0 stub. Real per-stage cost, MLIP surrogate, and on-chain attribution ship in Phase 2.
One sentence. The capability the ANN exposes, and the one thing it does really well.
Score a prediction 0–1.0 against literature + historical data, with citations.
The numbers the marketplace uses to rank this ANN. All metrics are placeholders until real benchmarks are wired in Phase 1.
Phase 0 stub benchmark
On a single CPU worker
Billed via the platform credit balance
Locked in Qearn to unlock access
Live metrics appear once deployed
Creator share + staker share + protocol fee
A realistic example input. The output you see on this page is from a Phase 0 stub backend. The format, latency, and cost are real, the data is illustrative.
{
"prompt": "Prediction: band_gap = 2.8 eV ± 0.4, material: LiNi0.5Mn0.5O2, source: mat-simulation-runner."
}{
"confidence": 0.72,
"literature_match": true,
"prior_measurements": 14,
"citations": [
"J. Electrochem. Soc. 2024",
"Nat. Mater. 2023"
]
}Real engine integration ships in Phase 2. The wire format here is what the real response will look like.
Three ways to use this ANN. Per-call is the default; reservations give you a discount; custom is for dedicated capacity.
Try the ANN with a 1,000-call reservation. No commitment.
Pre-purchased capacity. 10% discount on per-call pricing.
Private deployment, dedicated worker, custom SLAs.
Anyone can use, modify, and redistribute. Free of charge.
Other ANNs in the marketplace that share tags with this one.
Plan a material discovery workflow from a research question.
Ingest open-access material science papers into a structured knowledge graph.
Run DFT, MD, or ML surrogate simulations for material properties.
Every revision is a separate ANN with its own benchmarks. The current version is what's served by default.
| Version | Date | Status | Notes |
|---|---|---|---|
| v1.0.0 | Aug 2, 2026 | current | Initial release (Phase 0 stub backend). |
| v0.9.0 | Jul 24, 2026 | beta | Internal preview. Used for 1,000-paper ingestion benchmark. |
| v0.5.0 | Jul 12, 2026 | deprecated | First preview. Performance baseline. |
Real version history wired in Phase 1. For now, this is a static manifest.
This is one of the 8 material science ANNs in the Aigarth marketplace. Stake to participate in the network, and in the revenue when a discovery ships.
Deploy is a Phase 0 stub. Real per-stage cost, MLIP surrogate, and on-chain attribution ship in Phase 2. The dashboard at localhost:4000/material-science is the source of truth for what is real today.