3 ms·Toward Training Superintelligent Software Agents Through Self-Play SWE-RL2 points by klipt 10mo ago