Zulip Chat Archive
Stream: lean4
Topic: Autoformalization
David Afonso Valente (Mar 01 2024 at 16:31):
Hi! I'm aware that autoformalization isn't the main topic of interest however it seems like it would be great to have a reliable way to transate NL to FP seen as we would be able to generate a lot more data to train a F2F prover given all of the NL proofs available. Another possible use case I can think of is this would allow us to formally check NL proofs which would also greatly benefit Inf2Inf solutions and potentialy aid a hybrid solution in which we take advantadge of LLMs' impresive reasoning abilities. I am curious to know other peoples opinion on this matter.
Mario Carneiro (Mar 02 2024 at 06:07):
You should probably check out #Machine Learning for Theorem Proving which has a bunch of discussions related to this
Last updated: May 02 2025 at 03:31 UTC