mertkayacs/jevalt-bench
Viewer • Updated • 3.12k • 2.01k • 1
Benchmarks and datasets
Decision-model evaluation and training data, executed AI-agent trajectories and village simulation videos. Read each card for splits and licenses.
Note Decision tests: 1,277 English, 1,019 Turkish and 820 German rows, plus 390 live requests. Tests share the training pipeline.
mertkayacs/tholos-trajectories
Viewer • Updated • 263 • 181Note 263 executed tool-use trajectories: 237 train, 26 validation. Public sample; most validation runs were used to train Tholos-2B.
mertkayacs/jevalt-data
Viewer • Updated • 26k • 389 • 1Note Probability-labeled training rows in 3 languages: English, Turkish and German. Train, calibration and validation; per-row licenses.
mertkayacs/emberwick-videos
Updated • 669Note Village simulation clips of 3 decision models. Includes a 103-second problem film. Demonstrations, no benchmark scores.
-
mertkayacs/tholos-bench
Viewer • Updated • 160 • 237