研究人员开发了LongStraw，这是一个旨在实现强化学习（RL）后训练的执行栈，能够处理超过200万Token的上下文长度，即使在固定的GPU限制下也能实现。该系统解决了推理上下文长度与当前RL后训练能力之间日益增长的差距。LongStraw通过优化共享提示的评估并选择性地保留必要的模型状态来减少计算负载。对Qwen3.6-27B和GLM-5.2模型的初步实现表明，能够处理高达446万Token的提示，确立了极长上下文的执行可行性。 AI
影响 能够训练具有显著更长上下文窗口的AI代理，有可能提高它们处理复杂、多轮交互和大文档的能力。
排序理由 该集群包含一篇详细介绍LLM训练新方法的论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2607.14952v1 Announce Type: new Abstract: A growing gap separates inference context lengths from RL post-training: inference systems are approaching million-token contexts, while post-training workloads often remain at 256K tokens or below and rely on length generalization …
A growing gap separates inference context lengths from RL post-training: inference systems are approaching million-token contexts, while post-training workloads often remain at 256K tokens or below and rely on length generalization at deployment. The gap is especially important f…
A growing gap separates inference context lengths from RL post-training: inference systems are approaching million-token contexts, while post-training workloads often remain at 256K tokens or below and rely on length generalization at deployment. The gap is especially important f…