List your company

Knowledge gap detection on GSM8K

83.6AccuracyGRADEpos
45.6455.49565.3575.205Apr 3, 2026

Evaluation results

MethodLinks
GRADEposBackbone=Llama3.1-8b2026.0483.689.3
GRADEposBackbone=Gemma2-9b2026.0477.682.1
GRADEpreBackbone=Gemma2-9b2026.0474.578.3
P-EntropyBackbone=Gemma2-9b2026.0473.979.4
Align-PBackbone=Gemma2-9b2026.0472.777.6
GRADEposBackbone=Qwen2.5-7b2026.0471.977.2
GRADEpreBackbone=Llama3.1-8b2026.0471.476.7
GRADEpreBackbone=Qwen2.5-7b2026.0469.982.8
Align-PBackbone=Llama3.1-8b2026.0469.276.6
Align-PBackbone=Qwen2.5-7b2026.0464.566.9
P-EntropyBackbone=Llama3.1-8b2026.0464.349.6
JudgeBackbone=Llama3.1-8b2026.0463.9—
P-EntropyBackbone=Qwen2.5-7b2026.0462.754.1
ICBackbone=Qwen2.5-7b2026.0460.363.7
ICBackbone=Llama3.1-8b2026.0451.863.9
JudgeBackbone=Gemma2-9b2026.0451.3—
ICBackbone=Gemma2-9b2026.0450.754.1
JudgeBackbone=Qwen2.5-7b2026.0447.1—