List your company

Mathematical Reasoning on GSM8K original and paraphrased (train)

33.4S_refQwen-Math
-1.03967.9014516.842525.78355May 21, 2026

Evaluation results

MethodLinks
Qwen-MathData Variant=Original, Metric Type=ACC(%)2026.0533.445.8———1
Qwen-MathData Variant=Paraphrased, Metric Type=ACC(%)2026.0533.442———0.996
Qwen-MathData Variant=Original, Metric Type=Con(%)2026.0533.2—45.8——1
Qwen-MathData Variant=Paraphrased, Metric Type=Con(%)2026.0533.2—41.8——0.997
DeepSeek-MathData Variant=Original, Metric Type=ACC(%)2026.0522.229.8———0.989
DeepSeek-MathData Variant=Paraphrased, Metric Type=ACC(%)2026.0522.227.6———0.951
DeepSeek-MathData Variant=Original, Metric Type=Con(%)2026.0521.6—30——0.997
DeepSeek-MathData Variant=Paraphrased, Metric Type=Con(%)2026.0521.6—26.6——0.92
Qwen-MathData Variant=Original, Metric Type=P_first2026.050.488——0.532—0.962
Qwen-MathData Variant=Paraphrased, Metric Type=P_first2026.050.488——0.532—0.94
Qwen-MathData Variant=Original, Metric Type=P_all2026.050.412———0.5110.998
Qwen-MathData Variant=Paraphrased, Metric Type=P_all2026.050.412———0.4980.998
DeepSeek-MathData Variant=Original, Metric Type=P_first2026.050.38——0.463—0.998
DeepSeek-MathData Variant=Paraphrased, Metric Type=P_first2026.050.38——0.459—0.998
DeepSeek-MathData Variant=Original, Metric Type=P_all2026.050.285———0.3950.998
DeepSeek-MathData Variant=Paraphrased, Metric Type=P_all2026.050.285———0.3810.998