List your company

General-domain capability retention on AIME-2024, MATH-500, GSM8K, HumanEval, MMLU

62.9AIME-2024 AccuracyBase
-2.51614.46731.4548.433May 16, 2026

Evaluation results

MethodLinks
BaseModel=Qwen3-4B-It2026.0562.994.292.68677.682.6
MIXSDModel=Qwen3-4B-It, lambda=0.32026.0553.192.291.987.271.679.2
MIXSDModel=Qwen3-4B-It, lambda=0.52026.0552.193.492.484.87379.1
MIXSDModel=Qwen3-4B-It, lambda=02026.0543.89192.887.272.977.5
OPSDModel=Qwen3-4B-It2026.0543.391.292.388.47377.7
OPSD-NLLModel=Qwen3-4B-It, n=1, T=02026.0538.89091.886.665.374.5
BaseModel=Qwen3-8B2026.052683.491.886.673.172.2
MIXSDModel=Qwen3-8B, lambda=0.32026.0521.278.290.382.360.766.6
MIXSDModel=Qwen3-8B, lambda=0.52026.0520.479.291.384.157.866.6
OPSDModel=Qwen3-8B2026.0515.676.690.886.656.865.3
MIXSDModel=Qwen3-8B, lambda=02026.0514.474.288.684.143.360.9
OPSD-NLLModel=Qwen3-8B, n=1, T=02026.0512.574.888.981.153.862.2
BaseModel=Qwen3-1.7B2026.051172.480.460.458.556.5
MIXSDModel=Qwen3-1.7B, lambda=0.52026.055.457.67553.736.345.6
SFTModel=Qwen3-8B2026.055.441.663.284.173.153.5
MIXSDModel=Qwen3-1.7B, lambda=0.32026.054.258.677.741.529.142.2
OPSDModel=Qwen3-1.7B2026.053.354.675.943.336.742.8
MIXSDModel=Qwen3-1.7B, lambda=02026.053.14970.139.628.438
OPSD-NLLModel=Qwen3-1.7B, n=1, T=02026.051.941.664.82833.934
SFTModel=Qwen3-1.7B2026.0502.23.48.51.83.2
SFTModel=Qwen3-4B-It2026.0502.23502816.6