2 ms·

This kind of complaint makes it look like you stopped at the title and didn't even bother with the abstract, which says this: > In this work, we show that the

This kind of complaint makes it look like you stopped at the title and didn't even bother with the abstract, which says this: > In this work, we show that the internal representations of LLMs encode much more information about truthfulness than previously recognized. We first discover that the truthfulness information is concentrated in specific tokens, and leveraging this property significantly enhances error detection performance. "LLMs encode information about truthfulness and leveraging how they encode it enhances error detection" is a meaningful, empirically testable statement.

youoy 2y ago
If I look at the comments of this post my conclusion is that using "true" and "know" is hiding what the actual result is from a lot of people. So people get stuck on meaningless discussions. Seeing this makes me conclude that it is a bad choice from the knowledge transfer point of view for the scientific community. It would be better to use more objective/less emotionally charged words (which is what I understand by scientific language). If I now talk about my personal experience, when I read this article I have to "translate" in my head every appearance of those words to something that I can work with and is objective. And I find that annoying.