Comment by chenzhekl

I don't work with large language models, but out of curiosity: mimo's score on DeepSWE keeps going up, so why are they stopping training early? Is it due to budget constraints?