List your company

Reasoning on Hebrew Reasoning Benchmarks Suite (Copa, ARC-AI2, HellaSwag, MMLU, GSM8K, Psychometric Psi)

93.3Copa (HE)Gemma-3-27B-IT
87.78889.21990.6592.081May 11, 2026

Evaluation results

MethodLinks
Gemma-3-27B-ITTraining Stage=SFT2026.0593.391.463.672.582.854.376.3
HebatronTraining Stage=SFT2026.0591.98858.968.483.352.573.8
DictaLM-3.0-ThinkingTraining Stage=SFT2026.058891.261.760.270.242.368.9