anakin87/fact-checking-rocks · Interesting tool, good work!

Ive been curious about the idea of applying ML to combat misinformation. This HF space i find interesting. I like how it doesn't say FAKE, TRUE like other tools. (i must clarify i don't expect these systems to replace real human fact checkers anytime soon).

The stats and terminology i prefer over the the other binary(T/F) systems as i watch this space its very refreshing!! :)

While i know "fact checking" is for the humans as of now (30th of April 2025), and the space title is a little misleading, (less than the others i see on HF i might add). The idea that we can use ML to check the "bias" in an argument (which is possible to do as bias is more a lower level NLP task than a "deep reasoning"/"connect the dots across sources" task), or COI(confict of intrest) checking of sources (harder, more nuance).

I think it would be cool if there were a suite of tools (with the starting premise being its for research, later refined and integrated into other things), which have specialised programs for specific narrow subsets of the "fact checking" process. In simple terms i think its possible for human "fact checkers" to augment certain parts of their process (as long we ensure there is reasonably oversight over said suite).

This technology could be used to catch bias in the fact checkers themselves (thus making genuine "fact checking" work more accurate) (assuming this is done locally and transparently).

I must say they won't be as directly useful at the "fact checkers" level (maybe research?), problems that have no answer, where the jury is out completely (science for example, does have blind spots in some areas), or where a resource has partial information (under "fact checking", we tend to disregard this if no complementary information can be attached from elsewhere, an LLM may not).