I’ve noticed recurring mistakes in ChatGPT’s handling of probability and statistics problems, particularly in questions involving distributions (binomial, Poisson, etc.), expectation values, and variance calculations. While the steps are usually clear, there are instances where the final answers are incorrect, possibly due to calculation errors or misunderstanding the underlying statistical concepts. Improving the model’s accuracy in solving these types of problems would greatly enhance its usefulness for academic and technical purposes.
Try and use Code Interpreter to perform any mathematical work.
You can do something like ask o1 the question and take advantage of it’s reasoning capabilities, and then ask gpt4o to run the information through code interpreter to validate it.
1 Like