Ab test/agent action shadow - #55
Closed
AlexBybye wants to merge 4 commits into
Closed
Conversation
AlexBybye
marked this pull request as ready for review
August 28, 2026 17:27
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
实跑证据
6d992cc0...:AB 分支最近一次成功后端运行81ec11ae...:master 最近一次成功后端运行输出表现
AB 输出的内容表现:
主要问题:
AB 的核心结论是:
master 输出的内容表现:
线性代数伴随矩阵求逆公式线性方程组解的判定主要问题:
master 的核心结论是:
两边直接对比
master 的表现
master 的回答更像“直接完成任务”:先根据考试大纲安排复习顺序,再补充行列式、矩阵、秩、线性方程组等核心公式,整体内容完整,且没有触发重试。
它的问题是仓库引用偏少,只有 2 条。回答虽然能用,但有一部分内容主要来自用户大纲和通用知识,历年卷证据支撑相对弱。语言也更口语化,出现了“学妹”“考完请你喝奶茶”等风格化表达,是否合适取决于你想不想保留这种陪伴式语气。
abtest 的表现
abtest 的回答结构更规整,明确按五个章节组织:
行列式 → 矩阵 → 秩与方程组 → 向量组 → 特征值与二次型它还把“公式记不住”转成了“记母公式、理解公式之间的关系”,教学组织比 master 更清晰。引用数量是 master 的两倍,说明 Agent 的额外动作确实让最终回答覆盖了更多仓库依据。
代价也很明确:
回答质量上的共同问题
两边都存在同一个问题:在“未覆盖内容”里重新粘贴了用户提供的大纲。这不是事实错误,但对用户价值不高,会让回答显得冗长,也会稀释真正有用的复习建议。
两边还都把“历年题没有按知识点完整归组”说得比较诚实,没有直接编造命题概率或“必考结论”,这一点是可靠的。
AB 实验结论
这次结果不是“abtest 全面胜出”,而是一个很清楚的取舍:
master更好。abtest更好。我的判断是:当前不建议直接把 abtest 当成默认方案合并回 master。master 更适合作为线上基线;abtest 的“引用增强”值得保留,但需要先解决重复输出、重试导致的延迟,以及为什么一次额外决策会显著增加输出 token 的问题。
这只是各一次运行,不能证明长期稳定性;但就这两次最新基线而言,结论已经足够明确:master 胜在效率,abtest 胜在证据覆盖,整体上 master 暂时更划算。