Benchmarks

Reproduce our results in the browser.

Login

Benchmarks

Choose a benchmark family. The trust path is separate from the product upload workflow so users can verify the numbers directly.

Reproduce Published Cards

Open a locked production benchmark replay and jump directly to the selected result.

DUD-E 0.897 87-target replay DEKOIS2 0.942 76-target replay LIT-PCBA 0.890 15-target replay LP+ Relative 0.814 Relative affinity replay LP+ Absolute 0.853 Absolute affinity replay

IsoScreen Rank Discrimination

DUD-E and DEKOIS2 browser reproduction for active/decoy ranking.

0.916 mean AUC
Open Reproduce Runner

LP+ Absolute Affinity

CASF-285 live curated-reference scoring power; novel held-out uploads are about rho 0.76.

0.959
Open Absolute Benchmarks

LP+ Relative Affinity

Law74 relative-affinity replay across blind lead-optimization targets.

0.814 mean rho
Open Relative Benchmarks

Affinity Benchmark Validation

Held-out and independent validation evidence for absolute & relative binding, with downloadable summaries.

ρ 0.74-0.90
View Validation & PDFs

PDBbind 5,327 OOF

Full broad-panel absolute-affinity benchmark from the locked 1056 product-mode CSV.

R 0.731
Open PDBbind Benchmark

Reference Point

Benchmark pages compare LatticeZero against the FEP+ reference range and show browser-run timing where the runner is available.