> Then, we apply another filter, keeping only the cases that contain indisputable evidence — for example, hundreds of lines copied right down to the last whitespace error. We have virtually eliminated false positives at this point.

My friends who teach used the same level of filtering in the past: They suspected a lot of students from copying from each other, but they only took action on the cases where it was so undeniably obvious that it would be an open and shut case.

In the era of LLMs, that degree of precision is completely gone. However, the other signal is amplified: They have more students than ever getting 100s on homework and completely failing any exams. In the past it seems that many of the plagiarizers at least learned something, however minimal, in the process of copying homework and then trying to turn it into something that looked less suspicious. The LLM users are so brazen that they just submit the prompt, copy the output, and call it a day.