This study aims to verify the reliability of peer evaluation during persuasive and informative speaking in university liberal arts speech classes. To this end, it presented 46 university students with persuasive and informative speaking assignments and conducted a peer evaluation during their presentations. Subsequently, the students’ evaluation results were analyzed by applying the multi-faceted Rasch model. The analysis focused on evaluator appropriateness, evaluator bias, and the appropriateness of criteria and evaluation standards. The results of the analysis showed that 6 evaluators demonstrated inappropriate evaluation behavior during the persuasive speaking evaluation, and 13 were confirmed as over-fit and unfit evaluators in the informative speaking evaluation. Further, a greater over-fit reaction was found in the informative speaking evaluation for severity variability and evaluator agreement. Moreover, differences were found in using evaluation standards. The most severe evaluation was for the contents of persuasive speaking, and the most severe reaction was found in the delivery of informative speaking. In addition, a tendency for over-fit evaluation was found in the organization criteria for informative speaking. Thus, the results of the research highlighted different evaluation tendencies from the same evaluator based on the genre of speech. Therefore, there is a need for evaluator training based on genre.