OpenAI: This time, we are posting the results in the GitHub code repository, along with the paper revisions and citation specifications. For future publication of results, we promise to further optimize the quality of papers, improve literature citations, mathematical expressions, and presentation of results to make them easier for academics to understand. A large amount of proof formalized code completed in Lean language is available in the repository. In order to ensure the transparency and openness of the research, the repository also disclosed more process details of this study: including 10 model inference summaries, computational resource consumption estimates based on ChatGPT Pro's computing power, and statistics on the model's attempt to solve the problem. The average computing power consumed by each outcome is roughly equivalent to the computing power required by ChatGPT Pro for three hours of deep thinking.

OpenAI: This time, we are posting the results in the GitHub code repository, along with the paper revisions and citation specifications. For future publication of results, we promise to further optimize the quality of papers, improve literature citations, mathematical expressions, and presentation of results to make them easier for academics to understand. A large amount of proof formalized code completed in Lean language is available in the repository. In order to ensure the transparency and openness of the research, the repository also disclosed more process details of this study: including 10 model inference summaries, computational resource consumption estimates based on ChatGPT Pro's computing power, and statistics on the model's attempt to solve the problem. The average computing power consumed by each outcome is roughly equivalent to the computing power required by ChatGPT Pro for three hours of deep thinking.