微软研究：多智能体共享进度提升ARC-AGI性能
Banger paper from Microsoft Research and colleagues.
展示了多智能体协作的新范式，在ARC和压缩任务上取得显著突破，对Agent架构设计有直接参考价值。
Banger paper from Microsoft Research and colleagues.
来自微软研究院及其同事的出色论文。
It studies the potential benefits of agents that share progress while they work.
该研究探讨了在协作过程中共享进展的智能体（agents）所带来的潜在益处。
(bookmark it)
（收藏此链接）
Communication is still a challenge with multi-agent systems.
在多智能体系统中，通信仍然是一个挑战。
In this setup, agents have no predefined roles and communicate via a shared directory.
在此设置中，智能体没有预定义的角色，而是通过共享目录进行通信。
They report that a team of k agents that write their findings to a shared directory matches the success rate of 4k agents working independently on ARC-AGI-3.
他们报告称，一个由 k 个智能体组成的团队，将其发现写入共享目录，其成功率与 4k 个独立工作的智能体在 ARC-AGI-3 上的成功率相当。
The gap grows with k, and teams reliably solve some tasks that no single agent solves.
随着 k 的增加，这种差距会扩大，且团队能够可靠地解决某些单个智能体无法解决的难题。
The same setup beat best@k on polyomino packing and exceeded the prior best-known score. On MNIST compression, a four-agent team wrote a 1,957-byte classifier with 99.4% test accuracy, smaller than the best-known human solution.
相同的设置在多格骨牌填充任务上超越了 best@k 方法，并超过了此前已知的最佳分数。在 MNIST 压缩任务中，一个四人智能体团队编写了一个大小为 1,957 字节的分类器，测试准确率达到 99.4%，优于目前已知的最佳人类解决方案。
Independent agents still do better when compute is tight or when there is no clear measure of progress, so the paper also tells you when it might be a good idea to skip communication.
当计算资源紧张或缺乏明确的进展衡量标准时，独立智能体的表现仍然更好，因此该论文也指出了何时可能适合跳过通信。
Paper: https://arxiv.org/abs/2609.21032
论文：https://arxiv.org/abs/2609.21032
Chat with Paper: https://academy.dair.ai/papers/scaling-discovery-through-test-time-communication-2609.21032
与论文对话：https://academy.dair.ai/papers/scaling-discovery-through-test-time-communication-2609.21032
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