Laya Release - Typed Decisions Accuracy: leaderboard

typed-decisions benchmark: 400 cases, 2,000 decisions, four workflows (measured by Laya on the root checkpoint, zero-shot) 2000 Soft accuracy, Brier and ECE are held (card and BENCHMARKS disagree in the third decimal). Metric: accuracy (+ score MAE secondary).

Metric: accuracy (+ score MAE secondary). Source: raw.githubusercontent.com. 1 model tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Laya (421M)0.362

Interactive version: theaggregate.ai/benchmark?slug=laya-release-typed-decisions-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-16.