OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction Data

Benchmark Model Rank Results
arithmetic-reasoning-on-gsm8kOpenMath2-Llama3.1-70B (majority@256)#12Accuracy: 96.0
arithmetic-reasoning-on-gsm8kOpenMath2-Llama3.1-70B#25Accuracy: 94.9
arithmetic-reasoning-on-gsm8kOpenMath2-Llama3.1-8B (majority@256)#34Accuracy: 94.1
arithmetic-reasoning-on-gsm8kOpenMath2-Llama3.1-8B#60Accuracy: 91.7
math-word-problem-solving-on-mathOpenMath2-Llama3.1-70B (majority@256)#18Accuracy: 79.6
math-word-problem-solving-on-mathOpenMath2-Llama3.1-8B (majority@256)#22Accuracy: 76.1
math-word-problem-solving-on-mathOpenMath2-Llama3.1-70B#27Accuracy: 71.9
math-word-problem-solving-on-mathOpenMath2-Llama3.1-8B#38Accuracy: 67.8