Questioning the Questions: Sustaining Self-Evolution in Reasoning Models

Benchmark Model Rank Results
mathematical-reasoning-on-minerva-mathR-Quest (Qwen3-4B-Base)#5Accuracy: 58.82
math-word-problem-solving-on-mathR-Quest (Qwen3-4B-Base)#20Accuracy: 77.6
multiple-choice-question-answering-mcqa-on-supergpqaR-Quest (Qwen3-4B-Base)#21Accuracy: 29.4
mathematical-reasoning-on-olympiadbenchR-Quest (Qwen3-4B-Base)#36Accuracy: 46.67
arithmetic-reasoning-on-gsm8kR-Quest (Qwen3-4B-Base)#45Accuracy: 92.57