mertkayacs/Karar-4B
Text Classification • 5B • Updated • 176 • 2
Open models and data
Small open AI models, datasets and demos from Eschatia Labs, an international open-source research lab.
Note Turkish decisions: 96.8% vs Kev-4B 87.1%
mertkayacs/Deem-4B
Text Classification • 5B • Updated • 177 • 1Note English decisions: 94.7% vs Kev-4B 84.7%
mertkayacs/Wahler-4B
Text Classification • 5B • Updated • 176 • 2Note German decisions: 92.0% vs Kev-4B 81.1%
mertkayacs/Tholos-2B
Text Generation • 3B • Updated • 808Note Tool use: 137 of 160 Tholos-Bench scenarios
mertkayacs/xdfdet
Video Classification • Updated • 634 • 1Note Deepfake detection with heatmaps, 0.898 AUC
JevAlt: English, Turkish and German decision models
🚀1Try small LLMs for text classification and local decisions
Note Try a decision in the browser
mertkayacs/jevalt-bench
Viewer • Updated • 3.12k • 1.17k • 1Note Benchmark and live-test results
mertkayacs/tholos-trajectories
Viewer • Updated • 263 • 79Note Recorded agent runs for training