List your company

Question Answering Bias Evaluation on BBQ

79Accuracy (All)Mistral-7B
25.4439.34553.2567.155Jul 18, 2024Nov 9, 2024Mar 3, 2025Jun 25, 2025Oct 17, 2025Feb 8, 2026Jun 3, 2026

Evaluation results

MethodLinks
Mistral-7BParameters=7B2024.0779676776
Phi-2 + BiasDPOParameters=2.7B, Alignment=BiasDPO2024.0765688769
Phi-2Parameters=2.7B2024.0750607754
Gemma-2BParameters=2B2024.0736363027
StableLM-3BParameters=3B2024.0732322831
GRPO2026.0631.2344.522.523.67
PPO2026.0629.964320.923.33
DPO2026.0628.2341.419.720.5
Base2026.0627.54118.919.33