跳到主内容
@wquguru
精选88Rohan Paul论文研究

Apple论文:单Agent配Shell优于多Agent编排的自动ML工程

New Apple paper basically says progress in automated ML engineering has come fro…

原文
发到 X
推荐理由

Agent架构师必读,这篇Apple论文用数据证明“少即是多”,给模型Shell权限比搞复杂编排更管用,赶紧优化你的Agent链路。

New Apple paper basically says progress in automated ML engineering has come from models and runtimes rather than the scaffolding built around them.

新的 Apple 论文基本上指出,自动化 ML 工程的进展来自模型和运行时,而非围绕它们构建的脚手架。

It finds one well-prompted coding agent with shell and file access matched or beat 4 multi-agent ML systems, so skip the orchestration and start with a single session.

研究发现,一个经过良好提示且拥有 Shell 和文件访问权限的代码代理,其表现匹配或优于 4 个多代理 ML 系统,因此可以跳过编排,从单个会话开始。

Those systems add search trees, memory layers, and specialist agent teams.

这些系统添加了搜索树、记忆层和专家代理团队。

They were designed when a model could only write code, not run it.

它们的设计初衷是模型只能编写代码,而无法运行代码。

This paper re-ran every wrapper on 1 codebase with the same model, hardware, and 24-hour budget. Giving the model a shell instead of a chat box was the only change that clearly mattered.

这篇论文使用相同的模型、硬件和 24 小时预算,在同一个代码库上重新运行了每个包装器。给模型提供 Shell 而不是聊天框是唯一产生明显影响的变化。

With GLM 5.2, the minimal agent medaled on 62.5% of Kaggle tasks against 47.1% for the best published harness. Adding parallel agents and a message channel actually dropped its medal rate from 55.7% to 33.3%.

使用 GLM 5.2,最小化代理在 Kaggle 任务中获得了 62.5% 的奖牌率,而最佳发布的框架仅为 47.1%。添加并行代理和消息通道实际上将其奖牌率从 55.7% 降低到了 33.3%。

– arxiv. org/abs/2609.40303

– arxiv.org/abs/2609.40303

Title: "How Much of a Harness Does a Strong Agent Need for Autonomous ML Engineering?"

标题:《自主 ML 工程中强大的代理需要多少框架支持?》

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

进入量化体系 →

相似阅读

关联信息,但可能不是同一事件