AWS 新论文量化 Agent 切换模型的“交接税”成本
Great new paper from AWS on agent handoff tax.
Great new paper from AWS on agent handoff tax.
AWS关于代理交接税的新论文非常出色。
If you build agents today, you need to understand the so-called handoff tax.
如果你今天在构建代理，你需要理解所谓的交接税。
(bookmark it)
（收藏它）
Escalating to a stronger model mid-run is usually the resort when a cheap agent stalls.
当廉价代理停滞不前时，中途升级到更强的模型通常是最后的办法。
New work from AWS AI Labs measures how much that switch actually costs.
AWS AI实验室的新研究衡量了这种切换实际带来的成本。
Coding agents run for dozens of model calls, so teams escalate when a weak model struggles and downshift once the hard reasoning is done. Every switch forces the receiving model to continue a trajectory another model wrote.
编码代理会进行数十次模型调用，因此当弱模型遇到困难时，团队会升级，而在完成困难推理后则降级。每次切换都迫使接收模型继续执行另一个模型编写的轨迹。
Across pairs of Claude and GPT models, full-trajectory escalation recovers less than half the quality gap between the weak and strong model while adding a substantial cost premium. The authors call that penalty the handoff tax. Downshifting lands at a much better cost-quality point.
在Claude和GPT模型的各种组合中，全轨迹升级恢复的弱模型与强模型之间的质量差距不到一半，同时增加了显著的额外成本。作者将这种代价称为交接税。降级则在成本与质量之间取得了更好的平衡点。
Cutting the weak model's trajectory information improves escalation quality, while removing the strong model's trajectory hurts downshift quality.
减少弱模型的轨迹信息可以提高升级质量，而移除强模型的轨迹则会损害降级质量。
Paper: https://arxiv.org/abs/2608.24358
论文链接：https://arxiv.org/abs/2608.24358
Track more trending AI papers in our academy: https://academy.dair.ai/
在我们的学院中追踪更多热门AI论文：https://academy.dair.ai/
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