List your company
ResearchFormerly WizWand

Mathematical Reasoning on GAOKAO-en

73Pass RateSED-SFT
31.08841.96952.8563.731Feb 7, 2026
Updated 8mo ago

Evaluation results

MethodLinks
SED-SFTBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=RL, Optimization Objective=SED-SFT2026.0273
CrossEntropyBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=RL, Optimization Objective=CrossEntropy2026.0272.5
SED-SFT w/o maskBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=RL, Optimization Objective=SED-SFT w/o mask2026.0271.4
GEMBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=RL, Optimization Objective=GEM2026.0270.6
DFTBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=RL, Optimization Objective=DFT2026.0270.6
DFTBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=SFT, Optimization Objective=DFT2026.0270.1
SED-SFT w/o maskBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=SFT, Optimization Objective=SED-SFT w/o mask2026.0259.7
CrossEntropyBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=SFT, Optimization Objective=CrossEntropy2026.0259.5
SED-SFTBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=SFT, Optimization Objective=SED-SFT2026.0257.7
GEMBackbone=Qwen2.5-Math-7B-Instruct, Training Phase=SFT, Optimization Objective=GEM2026.0256.6
SED-SFTBackbone=Llama-3.2-3B-Instruct, Training Phase=RL, Optimization Objective=SED-SFT2026.0253
CrossEntropyBackbone=Llama-3.2-3B-Instruct, Training Phase=RL, Optimization Objective=CrossEntropy2026.0247.8
GEMBackbone=Llama-3.2-3B-Instruct, Training Phase=RL, Optimization Objective=GEM2026.0244.9
DFTBackbone=Llama-3.2-3B-Instruct, Training Phase=RL, Optimization Objective=DFT2026.0241
SED-SFT w/o maskBackbone=Llama-3.2-3B-Instruct, Training Phase=RL, Optimization Objective=SED-SFT w/o mask2026.0241
DFTBackbone=Llama-3.2-3B-Instruct, Training Phase=SFT, Optimization Objective=DFT2026.0237.1
CrossEntropyBackbone=Llama-3.2-3B-Instruct, Training Phase=SFT, Optimization Objective=CrossEntropy2026.0234.5
GEMBackbone=Llama-3.2-3B-Instruct, Training Phase=SFT, Optimization Objective=GEM2026.0233.8
SED-SFT w/o maskBackbone=Llama-3.2-3B-Instruct, Training Phase=SFT, Optimization Objective=SED-SFT w/o mask2026.0233
SED-SFTBackbone=Llama-3.2-3B-Instruct, Training Phase=SFT, Optimization Objective=SED-SFT2026.0232.7