PatentBind
Bigger is Better
Property Baseline
Ranks by molecular weight.

A property-based baseline that scores molecules solely by molecular weight. Larger molecules tend to have more intermolecular contacts, so this tests whether models beat the trivial heuristic of size-based ranking.

Design Rationale

If a model can't outperform 'bigger is better', it may be learning size artifacts rather than genuine binding signals.

Evaluation Scores

AUPRC0.1852

Area under the precision-recall curve. More informative when class balance is skewed.

AUROC0.3050

Area under the ROC curve. Measures discrimination ability across all thresholds.

EF 1%0.0000

Enrichment factor at 1%. How many actives found in the top 1% vs random.

EF 5%0.1876

Enrichment factor at 5%. How many actives found in the top 5% vs random.

Adjacent Accuracy0.5710

Fraction within ±1 rank of the true rank. Reflects triage decisions.

Concordance Index0.5959

Harrell's C statistic — probability that a random pair is correctly ordered.

Kendall τ0.2015

Rank correlation averaged within assay groups.

Exact Accuracy0.2189

Fraction of predictions matching the exact ordinal rank.

Pairwise Accuracy0.6350

Fraction of pairs correctly ordered by ordinal rank.

mrr0.3598
top1_accuracy0.0798
mrr0.5822
top1_accuracy0.3488
Compare with other models