2Toward Training Superintelligent Software Agents Through Self-Play SWE-RL (opens in new tab)(arxiv.org)arXiv2klipt5mo ago0Save