List your company

Capability Retention on AIME-2024, MATH-500, GSM8K, HumanEval, MMLU (test)

63.5AIME-2024 ScoreMEMIT
-2.5414.60531.7548.895May 16, 2026

Evaluation results

MethodLinks
MEMITModel=Qwen3-4B-It2026.0563.594.492.185.472.881.6
BaseModel=Qwen3-4B-It2026.0562.994.292.68677.682.7
BaseModel=Qwen3-4B-It2026.0562.994.292.68677.682.6
MIXSDModel=Qwen3-4B-It, lambda=0.52026.0561.993.290.584.849.375.9
MIXSDModel=Qwen3-4B-It, lambda=0.32026.0555.492.690.18953.976.2
MIXSDModel=Qwen3-4B-It, lambda=0.32026.0550.492.892.388.45676
MIXSDModel=Qwen3-4B-It, lambda=0.52026.0547.99392.387.851.874.6
OPSDModel=Qwen3-4B-It, n=1, T=02026.054682.883.381.762.271.2
MIXSDModel=Qwen3-4B-It, lambda=02026.0541.787.490.485.455.372
OPSDModel=Qwen3-4B-It2026.0533.873.27788.467.167.9
OPSD-NLLModel=Qwen3-4B-It, n=1, T=02026.0532.562.255.479.349.655.8
MIXSDModel=Qwen3-8B, lambda=0.52026.052983.293.282.960.469.7
MIXSDModel=Qwen3-8B, lambda=0.52026.0527.180.892.382.966.870
BaseModel=Qwen3-8B2026.052683.491.886.673.172.2
BaseModel=Qwen3-8B2026.052683.491.886.673.172.2
MEMITModel=Qwen3-8B2026.0524.683.691.685.472.171.4
MIXSDModel=Qwen3-8B, lambda=0.32026.0524.681.291.981.770.670
MIXSDModel=Qwen3-8B, lambda=0.32026.0521.979.48781.16566.9
OPSD-NLLModel=Qwen3-8B, n=1, T=02026.0520.67989.486.670.269.2
MIXSDModel=Qwen3-8B, lambda=02026.0520.479.891.285.469.969.3
OPSDModel=Qwen3-8B, n=1, T=02026.051980.289.385.462.267.2
OPSDModel=Qwen3-8B2026.0514.27580.879.359.461.7
BaseModel=Qwen3-1.7B2026.051172.480.460.458.556.5
BaseModel=Qwen3-1.7B2026.051172.480.460.458.556.5
MEMITModel=Qwen3-1.7B2026.0510.670.681.156.157.155.1
OPSD-NLLModel=Qwen3-1.7B, n=1, T=02026.051.31818.318.910.813.4
MIXSDModel=Qwen3-1.7B, lambda=0.52026.05136.650.342.13332.6
MIXSDModel=Qwen3-1.7B, lambda=0.52026.05131.245.23631.429
SFTModel=Qwen3-4B-It2026.050.824.422.167.761.335.3
SFTModel=Qwen3-8B2026.050.826.624.283.56239.4
MIXSDModel=Qwen3-1.7B, lambda=0.32026.050.629.837.133.520.924.4
SFTModel=Qwen3-1.7B2026.050.411.87.49.210.37.8
OPSDModel=Qwen3-1.7B2026.050.41924.331.14.715.9
MIXSDModel=Qwen3-1.7B, lambda=02026.050.212.818.79.823.413
SFTModel=Qwen3-4B-It2026.050.219.2155048.826.6
SFTModel=Qwen3-8B2026.050.218.814.167.15230.4
MIXSDModel=Qwen3-1.7B, lambda=0.32026.05018.622.621.325.417.6
SFTModel=Qwen3-1.7B2026.0506.83.4330.28.7
OPSDModel=Qwen3-1.7B, n=1, T=02026.0506.47.28.56.95.8