ArchAgent v2：通过分层约束优化实现CPU缓存预取突破
New from Google, DeepMind and Berkeley,
New from Google, DeepMind and Berkeley,
来自 Google、DeepMind 和 Berkeley 的最新成果，
This is a good example of why better agent architecture matters:
这是一个很好的例子，说明了为什么更优的智能体架构至关重要：
ArchAgent v2 shows a useful pattern for AI discovery: when a problem is too large to search at once, split it up and make the agent obey the same constraints as the final system.
ArchAgent v2 展示了 AI 发现的一个有用模式：当问题太大而无法一次性搜索时，将其拆分，并让智能体遵守与最终系统相同的约束条件。
Here, the problem is CPU cache prefetching: predicting which data the processor will need next.
在这里，问题是 CPU 缓存预取：预测处理器接下来需要哪些数据。
Searching L1D, L2, and LLC together was too slow and messy, so ArchAgent v2 optimized the cache levels in stages, kept the best designs, then refined them together.
同时搜索 L1D、L2 和 LLC（最后一级缓存）太慢且混乱，因此 ArchAgent v2 分阶段优化缓存层级，保留最佳设计，然后一起进行细化。
It also rejected candidates that exceeded the actual storage budgets.
它还拒绝了超出实际存储预算的候选方案。
With that setup, it beat the human-designed DPC4 champion.
通过这种设置，它击败了人类设计的 DPC4 冠军方案。
Its final design delivered a 3.8% IPC improvement over the competition baseline and edged BertiGO by 0.3%. On low-bandwidth single-core systems, it reached 4.6% versus BertiGO's 2.6%.
其最终设计相比竞赛基线实现了 3.8% 的 IPC（每指令周期数）提升，并以 0.3% 的优势领先 BertiGO。在低带宽单核系统上，它达到了 4.6%，而 BertiGO 为 2.6%。
Multi-core performance still lagged because those simulations make search much slower.
多核性能仍然落后，因为那些模拟使搜索速度大大减慢。
The takeaway for AI builders: better discovery agents need better search structure and real engineering constraints, not just a stronger model.
对 AI 开发者的启示：更好的发现智能体需要更优的搜索结构和真实的工程约束，而不仅仅是更强的模型。
– arxiv. org/abs/2608.09874
– arxiv.org/abs/2608.09874
Title: "ArchAgent v2: A Case Study with the Data Prefetching Championship"
标题：《ArchAgent v2：以数据预取锦标赛为例的案例研究》
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