What you're looking at
This is a read-only snapshot of the pipeline's output: the top ten candidates in each of the 80 categories, ranked by weighted score. The live pipeline runs offline against the full dataset — the numbers below are its published result, not a query being executed now.
- Score — how well the material's properties match what the application needs.
- Viability — a multiplier for cost, elemental abundance, supply-chain risk, and hazard.
- CLscore — synthesizability, from the KAIST Synthesizability-PU-CGCNN graph network.
- Weighted — score × viability. This is the ranking column.
Loading candidates…
Known limitations
Every number here is a prediction, not a measurement. Scores encode materials-science heuristics and GNN outputs, so a high rank means "worth a closer look," not "this works." The July 2026 audit documents the data defects found along the way, including a score-saturation bug that had collapsed eight categories into unranked ties, and the ALIGNN property columns that came back entirely null because the run had nothing to import. Both are written up rather than quietly fixed, because the failures are more instructive than the rankings.