Abstract: The widespread adoption of Machine Learning (ML) across industries has facilitated the use of data-driven decision-making and automation. However, concerns regarding the reliability and robustness of ML models persist. To ensure that ML models perform as intended, are unbiased, and generalize well to new data, comprehensive testing is essential. In this paper, Firstly, we elucidate and expound upon the obstacles that necessitate attention when assessing ML programs. Subsequently, we document the extant resolutions discovered in scholarly works pertaining to the assessment of ML programs. Lastly, we discern areas of deficiency within the literature concerning the evaluation of ML programs and proffer suggestions for prospective avenues of research within the scientific community.