提示词设计即工作设计：编码Agent的浪费与停止规则
Really useful revelation on prompting coding agent in this paper.
做Agent提示词工程的同学必看，这篇给出了可量化的浪费数据和具体的停止规则，能直接帮你优化提示词、省算力。
Really useful revelation on prompting coding agent in this paper.
这篇论文中对编程代理提示词的揭示非常有用。
Your coding-agent prompt may be spending compute on work you never needed.
你的编程代理提示词可能会在从未需要的工作上消耗计算资源。
A coding agent can solve the same bug correctly while doing far more work, just because the prompt asks it to consider several approaches.
一个编程代理可以正确解决同一个错误，但做了更多的工作，仅仅因为提示词要求它考虑多种方法。
Across six open-weight reasoning models on the frozen holdout, that instruction increased reasoning by 2.4–7.4× with no measured success gain.
在冻结的保留集上，六个开放权重推理模型中，该指令使推理量增加了2.4–7.4倍，但成功率没有可测量的提升。
The traces explain why: agents elaborated roughly three alternatives, discarded them, then implemented exactly one.
追踪记录解释了原因：代理详细阐述了大约三种备选方案，然后丢弃它们，最后只实现一种。
“Be absolutely certain” created a different failure mode, triggering repeated tests and extra checking after the fix looked done.
“绝对确定”造成了另一种失败模式，在修复看似完成后触发重复测试和额外检查。
At the highest observed redundant-verification level, runs cost 18.25× the clean-run median, used 15 versus 6 tool calls, and took 3× longer, while success stayed flat.
在观察到的最高冗余验证级别，运行成本是干净运行中位数的18.25倍，工具调用次数从6次增加到15次，耗时延长3倍，而成功率保持不变。
Prompt length barely mattered. Verbose repetition stayed around 1.0×.
提示词长度几乎无关紧要。冗长的重复保持在1.0倍左右。
A bounded instruction, start with the failing test, make the smallest sufficient change, run relevant tests, then stop, stayed at or below baseline reasoning while preserving diagnosis and validation.
一个有界指令——从失败的测试开始，做最小的足够改动，运行相关测试，然后停止——保持在基线推理水平或以下，同时保留了诊断和验证。
The tasks were small, at most four files, so this may not carry over to large architectural work.
任务很小，最多四个文件，因此这可能不适用于大型架构工作。
For coding agents, prompt design is work design: specify the useful work and give the agent a concrete stopping rule.
对于编程代理，提示词设计就是工作设计：指定有用的工作，并给代理一个具体的停止规则。
– arxiv. org/abs/2608.01347
– arxiv.org/abs/2608.01347
Title: "Same Task, Different Work: Prompt-Induced Waste in Coding Agents"
标题：“相同任务，不同工作：编程代理中提示词引发的浪费”
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