Every model (incl. Jev) we tested inflates security finding severity

Testing of various models, including Jev, revealed that they all tend to overstate the severity of security findings. The results of this testing were documented in a blog post and discussed on a news forum. The blog post is available for further reading, and a comments section is also available for discussion. The testing was conducted to benchmark the models' performance, specifically with regards to the Common Vulnerability Scoring System (CVSS). The findings suggest a potential issue with the accuracy of security severity ratings.