List your company

Reasoning and Math Problem Solving on BoolQ, COPA, WinoGrande, SVAMP, MAWPS (test)

88.2BoolQ AccuracyAUSteer-FFN
69.812874.586479.3684.1336Feb 4, 2026

Evaluation results

MethodLinks
AUSteer-FFNModel=Qwen3-8B, #Acts=<= 1002026.0288.290.867.2571.6789.5881.5
CAAModel=Qwen3-8B, #Acts=40962026.0287.8391.264.0171.6788.4580.63
AUSteer-HeadModel=Qwen3-8B, #Acts=<= 1002026.0287.7195.465.3576.3391.2783.21
SADIModel=Qwen3-8B, #Acts=kh * 1282026.0287.6593.264.257385.9280.8
VanillaModel=Qwen3-8B, #Acts=02026.0287.438963.6168.6784.5178.64
ITIModel=Qwen3-8B, #Acts=kh * 1282026.0287.490.865.5969.6787.3280.16
AUSteer-FFNModel=Gemma2-9B-it, #Acts=<= 1002026.0287.2597.670.887094.0883.96
STAModel=Gemma2-9B-it, #Acts=35842026.0287.0391.669.6168.6792.3981.86
CAAModel=Gemma2-9B-it, #Acts=35842026.0286.8589.468.356892.9681.11
ITIModel=Gemma2-9B-it, #Acts=kh * 2562026.0286.8288.468.116993.5281.17
SADIModel=Gemma2-9B-it, #Acts=kh * 2562026.0286.7992.469.536893.5282.05
AUSteer-HeadModel=Gemma2-9B-it, #Acts=<= 1002026.0286.799170.727194.6582.83
VanillaModel=Gemma2-9B-it, #Acts=02026.0286.6477.467.2567.6792.1178.21
AUSteer-HeadModel=LLaMA2-7B-Chat, #Acts=<= 1002026.0276.2775.452.837.6756.0659.64
AUSteer-FFNModel=LLaMA2-7B-Chat, #Acts=<= 1002026.0275.5782.853.283758.0361.34
CAAModel=LLaMA2-7B-Chat, #Acts=40962026.0274.9875.252.6434.3355.2158.47
SADIModel=LLaMA2-7B-Chat, #Acts=kh * 1282026.0274.3578.853.0436.3354.9359.49
ITIModel=LLaMA2-7B-Chat, #Acts=kh * 1282026.0274.177.252.836.6754.0858.97
VanillaModel=LLaMA2-7B-Chat, #Acts=02026.0270.5270.850.913651.8356.01