Training capable coding agents via reinforcement learning (RL) requires diverse tasks with reliable verifiers. Open-source codebases offer a rich source of s…
机构:小米
来源:arXiv 2609.22068 | AI4Papers 论文推荐平台