Toward Training Superintelligent Software Agents Through Self-Play SWE-RL (Meta)

1 points| xhevahir | 2 months ago |arxiv.org