Post #1998269
2026-05-04 15:47 UTC
We identified and classified the documents we studied based on a well-designed prompt to an LLM. Then to ensure robustness we employed a different LLM as an independent validator.
Replies (3)
-
@nyhan@fediscience.org 2026-05-04 15:49
BTW based on the fact that the appendix is linked right after the phrase "well-designed prompt" you might expect that the appendix talks about how the prompt was created, iterated, tested -- but no. The appendix has the prompt (which is better than nothing) but nothing about why the authors are so confident that it is designed well.
-
@Heidentweet@todon.eu 2026-05-04 15:50
@nyhan@fediscience.org 😬 and
-
@alxndr@tech.lgbt 2026-05-04 16:50
@nyhan@fediscience.org 🫠 ~~pLeASe bE VErY tHOrouGH, pLEAse!!1~~