jietang· @jietang · X·· 2026-07-14AI 评分47
AI 导读
我们的单轮异步优化（SAO）能够稳定训练一千步，并在智能体编码与推理基准（如SWE-Bench Verified、BeyondAIME和IMOAnswerBench）上持续优于GRPO及其变体。https://arxiv.org/pdf/2607.07508
正文 · 原文
Our Single-rollout Asynchronous Optimization (SAO), is able to train stably for one thousand steps and consistently outperform GRPO and its variants on agentic coding and reasoning benchmarks,
such as SWE-Bench Verified, BeyondAIME, and IMOAnswerBench. https://arxiv.org/pdf/2607.07508
来源：jietang · x.com