List your company

Scientific Reasoning on GPQA (test) (Accuracy and Length Metrics)

0.609AccuracyCAT
0.329240.401870.47450.54713Jul 1, 2026

Evaluation results

MethodLinks
CATBackbone=Qwen3-8B2026.070.6094,8194,3930.3340.309
Qwen3-8BBackbone=Qwen3-8B2026.070.5997,2416,354——
DASTBackbone=Qwen3-8B2026.070.5996,3805,5230.1190.131
ConCISESimPOBackbone=Qwen3-8B2026.070.5863,9393,4640.4560.455
OverThinkSimPOBackbone=Qwen3-8B2026.070.5713,7123,3890.4870.467
CATBackbone=DeepSeek-R1-Distill-Qwen-7B2026.070.547,7816,4250.090.13
DASTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.070.528,0687,0210.0570.049
ConCISESimPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.070.5026,0305,0110.2950.321
OverThinkSimPOBackbone=DeepSeek-R1-Distill-Qwen-7B2026.070.4987,2705,8560.150.207
R1-7BBackbone=DeepSeek-R1-Distill-Qwen-7B2026.070.4928,5537,382——
CATBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.070.379,7168,3480.0380.097
DASTBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.070.3679,9808,9930.0120.027
OverThinkSimPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.070.3578,9087,9670.1180.138
ConCISESimPOBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.070.3497,8906,9950.2190.243
R1-1.5BBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.070.3410,1009,240——