List your company

Question Answering on Xiezhi (test)

66.4Score (History)KaFT
48.92853.4645862.536May 21, 2025

Evaluation results

MethodLinks
KaFTBackbone=LLaMA3-8B, Training Domain=History2025.0566.46667.666.67
KaFTBackbone=LLaMA3-8B, Training Domain=Law2025.056665.661.264.27
Vanilla SFTBackbone=LLaMA3-8B, Training Domain=History2025.0564.46667.265.87
Vanilla SFTBackbone=LLaMA3-8B, Training Domain=Law2025.0564.464.460.863.2
No-conflictBackbone=LLaMA3-8B, Training Domain=Law2025.0561.6646061.87
No-conflictBackbone=LLaMA3-8B, Training Domain=History2025.056066.865.264
KaFTBackbone=LLaMA3-8B, Training Domain=Engineering2025.0558.867.256.860.93
Vanilla SFTBackbone=LLaMA3-8B, Training Domain=Engineering2025.0557.665.256.859.87
Self-aligningBackbone=LLaMA3-8B, Training Domain=Law2025.0557.664.45860
Self-aligningBackbone=LLaMA3-8B, Training Domain=History2025.0556.86660.461.07
No-conflictBackbone=LLaMA3-8B, Training Domain=Engineering2025.0556.865.25659.33
Self-aligningBackbone=LLaMA3-8B, Training Domain=Engineering2025.0552.863.655.257.2
BaseBackbone=LLaMA3-8B, Training Domain=History2025.0549.659.251.653.47
BaseBackbone=LLaMA3-8B, Training Domain=Engineering2025.0549.659.251.653.47
BaseBackbone=LLaMA3-8B, Training Domain=Law2025.0549.659.251.653.47