A new article notes that journal articles reporting how well machine learning models solve certain kinds of equations are often overly optimistic. The researchers suggest two rules for reporting results and systemic changes to encourage clarity and accuracy in reporting.
