CoreWeave RL Rollouts 免重部署加载权重，训练循环提速约 15 倍
CoreWeave 出了个 RL Rollouts，训练时换 checkpoint 不用重新部署，快了 15 倍，NVIDIA 用它把 BrowseComp 从 37% 拉到 45%。
CoreWeave 发布 RL Rollouts，把强化学习训练中的新 checkpoint 权重直接加载到在线部署里，不打断在途请求，比传统重部署周期快约 15 倍。NVIDIA 和 You.com 用它对 Nemotron 3.5 Lightning 做后训练以适配网页搜索。结果是 BrowseComp 准确率从 36.97% 提升到 45.45%，工具调用次数减少 30.24%。
In reinforcement learning, inference is part of the training loop. Every checkpoint used to mean a redeploy, and the trainer waited.
CoreWeave RL Rollouts load the new weights into a live deployment without touching in-flight requests, about 15x faster than a redeploy cycle.
@nvidia and @youdotcom used it to post-train Nemotron 3.5 Lightning for web search, lifting BrowseComp accuracy from 36.97% to 45.45% while reducing tool calls by 30.24%.
Full breakdown here: https://t.co/cTcqSRGeCB
