The researchers found a new way to compare the averages of two groups when we don't know their variances and they might not be the same. They created a test called the adjusted likelihood ratio test, which works well in terms of accuracy and power compared to other tests. This new test is closely related to the most commonly used test for this problem.