Comment by Tiberium

I have a suspicion that this paper is basically a summary with some benchmarks, done with LLMs.