This is extremely interesting, because while I'd never heard of the '2 sigma problem' [1] before, one university class I had seems to have been largely modeled on it, but with a very different angle. It was a 'self paced' electrical engineering course where we were given a textbook and free to advance through it at our own pace - kind of farcically, since you needed to complete at least 2 chapters per week to finish by the end of the semester.

Moving forward to the next chapter required, exactly as described in that paper, the completion of a problem set and then a score of at least 90% on a test demonstrating mastery of the previous chapter, sometimes accompanied by also demonstrating that skill in a lab. But far from 1 on 1, this entire class was effectively 0 on infinity. The teaching assistant/proctors that we engaged with were there only to grade your work and provided minimal feedback.

And indeed it was one of the most educational 'classes' I ever took. But I think this challenges the concept that it has anything to do with 1 on 1 attention. But rather the outcome seems practically tautological - a good way to get people to perform to the point of mastery is to require that they perform to the point of mastery. Of course, at scale, all you're really doing is weeding out the people that are unable to achieve mastery. And indeed that class was considered a weed out course.

[1] - https://en.wikipedia.org/wiki/Bloom's_2_sigma_problem