List your company

Logical Reasoning on HRLs (High-Resource Languages)

93.4BPA ScoreGemma-2-9B
27.46444.58261.778.818Feb 2, 2026

Evaluation results

MethodLinks
Gemma-2-9BConfiguration=SFT2026.0293.44.4
Gemma-3-12BConfiguration=Steering2026.0290.1—
Gemma-2-9BConfiguration=Steering2026.0289—
Qwen-3-14BConfiguration=Steering2026.0286.9—
Qwen-2.5-7BConfiguration=Steering2026.0286.1—
Ministral-14BConfiguration=Steering2026.0286.1—
Qwen-2.5-7BConfiguration=SFT2026.0284.6-1.5
Mistral-7BConfiguration=SFT2026.0278.22.8
Mistral-7BConfiguration=Steering2026.0275.4—
Gemma-3-12BConfiguration=Base Model2026.0263.5—
Gemma-2-9BConfiguration=CoT2026.0258.5-30.5
Qwen-2.5-7BConfiguration=CoT2026.0257.7-28.4
Qwen-3-14BConfiguration=Base Model2026.0257.5—
Ministral-14BConfiguration=Base Model2026.0249.2—
Gemma-2-9BConfiguration=Base Model2026.0243.9—
Qwen-2.5-7BConfiguration=Base Model2026.0242.9—
Mistral-7BConfiguration=CoT2026.0234.2-41.2
Mistral-7BConfiguration=Base Model2026.0230—