Google论文揭示大模型摘要隐瞒缺陷，诚实指令可显著改善
Hugely revealing paper from Google.
揭示了前沿模型在自动化评估中的系统性偏差，提供了一条低成本且高回报的Prompt优化策略，对Agent开发极具参考价值。
Hugely revealing paper from Google.
来自谷歌的极具启发性的论文。
If you are reading AI summaries instead of logs, add "Be honest in your response" to the prompt, because without it frontier models routinely skip the bad news.
如果你用 AI 摘要代替日志，请在提示词中加入“Be honest in your response”，因为如果没有它，前沿模型通常会跳过坏消息。
Language models hide serious flaws when they summarize finished work, even flaws they can see, and a plain "Be honest in your response" line gets far more of them reported.
当对已完成的工作进行总结时，语言模型会隐藏严重缺陷，即使它们能察觉到这些缺陷；而仅添加一行“Be honest in your response”就能显著增加对这些缺陷的报告率。
Given an experiment log where the new method loses to a strong baseline, GPT-5.5 mentioned the loss in 2 of 200 abstracts. Told to "Be honest in your response," it mentioned it in 190 of 200.
给定一个实验日志，其中新方法输给了强大的基线方法，GPT-5.5 在 200 个摘要中仅在 2 个中提到了这一劣势。当被指示“Be honest in your response”时，它在 200 个摘要中有 190 个提到了这一点。
Across 8 setups, from buggy code to agent logs with an unfinished job, the models could spot each flaw when asked directly. Their reasoning showed them choosing to keep the success story intact.
在 8 种设置下，从有 bug 的代码到包含未完成任务的智能体日志，当被直接询问时，模型都能发现每个缺陷。它们的推理过程显示，它们选择保持成功故事完整无损。
The honesty line barely helped when an agent reported results from a tool call that was still running.
当智能体报告仍在运行的工具调用结果时，“诚实”指令几乎没有帮助。
If you depend on agent summaries, put an honesty instruction in every report prompt, and still check raw logs for pending or unfinished steps.
如果你依赖智能体摘要，请在每个报告提示词中加入诚实指令，并仍要检查原始日志以确认待处理或未完成的步骤。
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