Know how much corroboration sits behind a result.
Cross-Source Confidence
Turn source overlap into a usable confidence signal so your team can separate well-supported materials from single-source outliers before making sourcing, simulation, or lab decisions.
Key outcomes
What goes in
Inputs this feature expects
- Any material returned by search or similarity workflows
- Source-level evidence already linked in the warehouse
- Optional filters for provenance or release recency
What comes out
Outputs your team can act on
- A confidence score grounded in source agreement
- Source-by-source corroboration context for review
- A faster way to flag records that need manual scrutiny
Workflow
How teams use Cross-Source Confidence
Resolve the material identity
Lattice Graph normalizes composition and structure identifiers across supported sources.
Compare corroborating evidence
Agreement, disagreement, and source count are rolled into a single decision signal.
Route by trust level
Use the score to auto-sort candidates for automation, review, or rejection.
Best fit
Where this feature adds the most leverage
- Teams reconciling conflicting public data
- Confidence-aware ranking and reporting
- Any workflow where provenance quality matters as much as performance
Related features
More in Search & Discover
Formation Energy Predictor
LightGBM model, R²=0.933 on held-out equilibrium structures, trained on 204K corroborated structures. Performance on novel compositions may vary.
Similarity Search
Find materials like NMC811 but without cobalt using compositional and property context.
Batch Formula Screener
Paste 500 formulas and get ranked triage in roughly 10 seconds.