Microsoft AI's Luca Soldaini acknowledged a terminology correction for the SWE-bench benchmark raised by academic Ofir Press · Digg