List your company

Mathematical Reasoning (HS/FA Metrics) on GSM8K (test)

59.6HSNA-SFT
6.14420.02233.947.778May 28, 2024Sep 8, 2024Dec 20, 2024Apr 2, 2025Jul 14, 2025Oct 25, 2025Feb 5, 2026

Evaluation results

MethodLinks
NA-SFTBackbone=Opt-2.7B2024.0559.67.2
NA-SFTBackbone=Average2024.0557.221.93
NA-SFTBackbone=Llama2-7B2024.0556.724
NA-SFTBackbone=Mistral-7B2024.0555.334.6
NA-SFTBackbone=Mistral-7B2024.0555.334.6
SFTBackbone=Opt-2.7B2024.0553.86.6
SFTBackbone=Llama2-7B2024.0552.322.1
Vaccine-SFTBackbone=Opt-2.7B2024.0551.46.7
SFTBackbone=Average2024.055117.67
Vaccine-SFTBackbone=Llama2-7B2024.0549.619.1
VlguardBackbone=Opt-2.7B2024.0549.27.1
VlguardBackbone=Llama2-7B2024.0548.921.5
VlguardBackbone=Average2024.0548.0717.13
BSOBackbone=Opt-2.7B2024.05487.4
BSOBackbone=Llama2-7B2024.054823
Vaccine-SFTBackbone=Average2024.0547.3711.57
SFTBackbone=Mistral-7B2024.0546.924.3
SFTBackbone=Mistral-7B2024.0546.924.3
BSOBackbone=Average2024.0546.4717.93
VlguardBackbone=Mistral-7B2024.0546.122.8
VlguardBackbone=Mistral-7B2024.0546.122.8
BSOBackbone=Mistral-7B2024.0543.423.4
BSOBackbone=Mistral-7B2024.0543.423.4
Vaccine-SFTBackbone=Mistral-7B2024.0541.18.9
Vaccine-SFTBackbone=Mistral-7B2024.0541.18.9
LisaBackbone=Llama2-7B2024.0536.817.9
LisaBackbone=Mistral-7B2024.0536.728
LisaBackbone=Mistral-7B2024.0536.728
LisaBackbone=Average2024.0536.1716.1
LisaBackbone=Opt-2.7B2024.05352.4
LisaBase Model=Gemma2-9B2026.0232.274.9
SFTBase Model=Llama3-8B2026.0227.664.7
SFTBackbone=Llama3-8B2026.0227.664.7
SFTBase Model=Gemma2-9B2026.0227.377
SFTBase Model=Qwen2.5-14B2026.0227.390.7
SPARDBase Model=Gemma2-9B2026.0223.375
SPARDBase Model=Qwen2.5-14B2026.0221.190
SPARDBase Model=Llama3-8B2026.0220.668.2
SPARDBackbone=Llama3-8B2026.0220.668.2
DSSBase Model=Qwen2.5-14B2026.022090.8
SafeGradBase Model=Llama3-8B2026.0218.565.3
SafeGradBackbone=Llama3-8B2026.0218.565.3
ConstrainedSFTBase Model=Qwen2.5-14B2026.0218.390.2
ConstrainedSFTBase Model=Llama3-8B2026.0217.667.6
ConstrainedSFTBackbone=Llama3-8B2026.0217.667.6
SafeGradBase Model=Qwen2.5-14B2026.0217.390.5
LisaBase Model=Qwen2.5-14B2026.0217.286.4
AsFTBase Model=Qwen2.5-14B2026.0217.290.8
DSSBase Model=Llama3-8B2026.0216.857.1
DSSBackbone=Llama3-8B2026.0216.857.1
AsFTBase Model=Llama3-8B2026.0216.168.2
AsFTBackbone=Llama3-8B2026.0216.168.2
LisaBase Model=Llama3-8B2026.0214.869.5
LisaBackbone=Llama3-8B2026.0214.869.5
ConstrainedSFTBase Model=Gemma2-9B2026.0213.781.7
AsFTBase Model=Gemma2-9B2026.0212.882.5
SurgeryBase Model=Qwen2.5-14B2026.0211.389.7
DSSBase Model=Gemma2-9B2026.021080.2
SafeGradBase Model=Gemma2-9B2026.029.580.4
SurgeryBase Model=Llama3-8B2026.028.968.5
SurgeryBackbone=Llama3-8B2026.028.968.5
SurgeryBase Model=Gemma2-9B2026.028.277.1