AgentZip：利用冗余压缩降低高并发Agent沙箱内存占用
Very cool paper on memory compression for agents.
Agent工程落地常受限于内存瓶颈，这篇论文给出了具体的系统级优化方案与实测数据，对大规模部署Agent有直接参考价值。
Very cool paper on memory compression for agents.
一篇关于智能体内存压缩的非常棒的论文。
If you run many agent sandboxes in parallel for RL or evals, memory becomes highly redundant. This work suggests that compressing against that redundancy cuts sandbox memory by up to 8.7x.
如果你并行运行许多用于强化学习（RL）或评估的智能体沙箱，内存会变得高度冗余。这项工作表明，针对这种冗余进行压缩可将沙箱内存减少多达 8.7 倍。
Memory is becoming the capacity limit for high-fanout agent workloads.
内存正成为高扇出智能体工作负载的性能瓶颈。
One task can spawn many concurrent sandboxes, and they all start from the same template and run related trajectories.
一个任务可以生成许多并发沙箱，它们都从相同的模板启动并运行相关的轨迹。
HKUST researchers measured 76 to 96% of pages with template-relative or cross-sandbox redundancy.
香港科技大学的研究人员测量发现，76% 到 96% 的页面存在基于模板的或跨沙箱的冗余。
AgentZip compresses pages against the template and against sibling sandboxes, including pages that are similar without being identical.
AgentZip 针对模板和兄弟沙箱对页面进行压缩，包括那些相似但不完全相同的页面。
It runs expensive compression while the agent is waiting on the LLM, and it prefetches pages at restore time to control slowdown.
它在智能体等待大语言模型（LLM）响应时执行昂贵的压缩操作，并在恢复时预取页面以控制性能下降。
Results:
结果：
Sandbox-owned memory drops by up to 8.7x, against 2.1x for the Linux configuration. Aggressive compression slows execution by 3.1x on its own, and the scheduling and prefetching bring that down to 1.40x.
沙箱占用的内存最多可减少 8.7 倍，而 Linux 配置下仅为 2.1 倍。激进压缩本身会使执行速度降低 3.1 倍，但通过调度和预取技术将其降至 1.40 倍。
Paper: https://arxiv.org/abs/2609.11294
论文：https://arxiv.org/abs/2609.11294
Chat with Paper: https://academy.dair.ai/papers/memory-compression-for-high-fanout-agent-sandboxes-2609.11294
与论文对话：https://academy.dair.ai/papers/memory-compression-for-high-fanout-agent-sandboxes-2609.11294
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力