List your company

Mathematical Reasoning on GSM8k 4 options (test)

50.3AccuracyKP
31.89236.67141.4546.229Sep 28, 2025

Evaluation results

MethodLinks
KPModel=Qwen 2.5 7B2025.0950.31000
KAPPAModel=Qwen 2.5 7B, Intervention layers=12025.0949.668.80.78
KAPPAModel=Qwen 2.5 7B, Intervention layers=62025.0949.266.30.76
KAPPAModel=Qwen 2.5 7B, Intervention layers=32025.0949.165.40.77
DoLAModel=Qwen 2.5 7B2025.0948.1600.86
BaseModel=Qwen 2.5 7B2025.0947.860.30.86
CAAModel=Qwen 2.5 7B2025.0947.760.30.85
KAPPAModel=Llama 3.1 8B, Intervention layers=62025.0936.675.90.27
KPModel=Llama 3.1 8B2025.0936.61000
KAPPAModel=Llama 3.1 8B, Intervention layers=12025.0934.973.80.37
KAPPAModel=Llama 3.1 8B, Intervention layers=32025.0934.666.20.27
DoLAModel=Llama 3.1 8B2025.0933.249.70.58
CAAModel=Llama 3.1 8B2025.0932.953.80.38
BaseModel=Llama 3.1 8B2025.0932.653.70.35