Status
successful_prediction
Median lDDT
0.9625
Targets
3 / 3
successful predictions and scores.
Run Boundary
This is a real inference smoke, not a publication-grade benchmark. The target set is deliberately tiny and classic, the recipe is an engineering fixture, and the comparison value is an internal regression value rather than a SimpleFold paper headline metric.
Per-Target Results
| Target | pLDDT mean | Wall clock | lDDT | TM-score | RMSD |
|---|---|---|---|---|---|
1L2Y | 97.5245 | 121.51 s | 0.9604 | 0.6603 | 0.43 |
1CRN | 91.1615 | 144.14 s | 0.9865 | 0.9707 | 0.37 |
1UBQ | 88.9500 | 118.22 s | 0.9625 | 0.9650 | 0.70 |
Verdict
The deterministic fixture verdict reports median lDDT 0.9624857567529903 versus fixture value 0.9694790633021089, a relative error of 0.7213468360316005%.
Artifacts
card.json
Machine-readable result card with predictions, scores, validation, and audit count.
predictions.json
Per-target prediction records and predicted PDB text.
scores.json
Per-target lDDT, TM-score, and RMSD.
verdict.json
Deterministic fixture validation verdict.
Next Gate
The next run should use the public Cloudflare Tunnel path through science.eigenatlas.com, deploy the updated orchestrator, and compare a no-fixture Reader extraction against the fixture recipe.