Skip to content
Agentmart
Submit a tool

method-ab-experiment

设计或审查模型、Agent、缓存、上下文、工具策略、runtime 优化的 A/B/qualification 实验时使用。用于锁唯一自变量、隔离 cache/session/evidence/model residency 等跨 run…

Skill by deyi2026

IndexedApache-2.0★ 1
Write a review

Chalk Labs

Need help deploying method-ab-experiment?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/deyi2026-method-ab-experiment

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

设计或审查模型、Agent、缓存、上下文、工具策略、runtime 优化的 A/B/qualification 实验时使用。用于锁唯一自变量、隔离 cache/session/evidence/model residency 等跨 run 状态、确认机制真正触发,并把“机制生效”“任务收益”“可推广/可发布”分层裁决;特别适合本地模型和长 Agent benchmark。简单确定性单测不需要加载。

Use cases: agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/method-ab-experiment

What's inside

Skill name
method-ab-experiment
Description
设计或审查模型、Agent、缓存、上下文、工具策略、runtime 优化的 A/B/qualification 实验时使用。用于锁唯一自变量、隔离 cache/session/evidence/model residency 等跨 run 状态、确认机制真正触发,并把“机制生效”“任务收益”“可推广/可发布”分层裁决;特别适合本地模型和长 Agent benchmark。简单确定性单测不需要加载。
License
Apache-2.0
Instructions
113 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# A/B Experiment Qualification

目标是获得可归因结论,不是让两组“都跑一次”。任何不能隔离的状态都应成为结论限制,而不是被忽略。

## 开始前先回答

1. 唯一自变量是什么?能否用一句话写清?
2. 除它之外,模型、prompt、tool schema、任务、runtime、参数、数据是否真的相同?
3. 哪些状态可能跨 run 泄漏:prompt cache、KV/cache residency、session、Evidence、文件、provider client、进程热身?
4. 哪些指标属于正确性、哪些属于结构效率、哪些受顺序影响只能观察?
5. 样本如何证明被测机制真的发生过?

## 实验前冻结

记录至少这些事实:

- exact model/provider;
- prompt/task 的可比标识;
- reasoning/temperature/max_tokens 等关键生成参数;
- 工具面及 schema 数量/版本;
- 数据目录/session 隔离方式;
- 服务进程与模型 residency;
- 被测 feature 的精确开关和值;
- 不能清除或不能随机化的共享状态。

不要为了“公平”而同时加载两个本地大模型;若机器资源会互相干扰,应顺序测试并把顺序效应写入限制。

## 默认路径

1. 建立 OFF/baseline,并确认任务能正常运行。
2. 建立 ON/candidate,只改变目标变量。
3. 在共同分叉点之前检查两组输入与行为是否保持同构;若早已不同,先解释差异来源。
4. 确认 ON 组确实触发目标机制,而不是 feature 开着但路径从未命中。
5. 收集结构指标和最终任务结果。
6. 对不可隔离的 warm cache/顺序效应,避免把 elapsed/cache-hit 单独当因果证据。
7. 加至少一个反例或 adversarial case,检查方法没有只适配主 benchmark。

## 指标分层

### 正确性

- 最终答案/任务验收;
- exact source/evidence recovery;
- 工具副作用是否正确;
- 是否出现新错误或遗漏。

### 行为结构

- rounds;
- tool calls / unique calls / duplicates;
- 无信息增量重复动作;
- recovery/hydration 次数;
- post-sufficiency extra rounds;
- provider-visible chars / tokens_in / prefill。

### 只能谨慎观察

- wall-clock elapsed;
- cache hit ratio;
- 首轮/后轮速度;
- 受模型 warm residency、OS 调度、共享 cache 影响的性能数值。

这些指标若未做随机顺序、冷启动或独立进程控制,不应独自支持强因果结论。

## 三层裁决

实验结束必须分开回答:

1. **Mechanism**:机制是否按设计真实触发?
2. **Task benefit**:触发后是否改善任务正确性或行为效率?
3. **Promotion**:证据是否足以推广到默认路径/公开 claim?

`Mechanism PASS` 不自动推出 `Task benefit PASS`;单个任务收益也不自动推出 promotion。

## 反例设计

至少选择一种与主样本不同的情况:

- 主样本需要 heavy path,反例只需 simple path;
- 主样本事实位于 head,反例把关键事实放 middle/tail;
- 主样本机制一定触发,反例故意保持在阈值以下;
- 主样本是长 Agent,反例是短任务,检查优化是否制造额外税。

## 停止/重做条件

以下情况不要继续解释结果,应先重做实验:

- A/B 实际用了不同模型或 prompt;
- 数据目录缺配置导致一组走 fallback;
- 被测机制根本未触发;
- 同时改变了多个关键变量且无法拆分贡献;
- 一组出现与实验无关的基础设施失败;
- 比较指标的统计口径不同。

## 反模式

- “ON 更快,所以机制更好”,但 ON 是第二个 warm run;
- feature flag 打开就视为机制已验证;
- 只测一个专门为方案设计的 happy path;
- 模型、prompt、tool schema 中有一项偷偷漂移;
- 把 cache hit、elapsed 的顺序效应包装成确定因果;
- 为得到好结果临时加入 prompt-visible completion hint,导致实验测的是另一件事。

## 方法边界

本方法提供实验纪律,不替模型决定哪些指标最重要,也不预设 ON 必须胜出。负结果同样是有效结论;发现 candidate 增加恢复税、固定偏差或收敛问题时,应如实 HOLD。

## 可沉淀内容

具体 run 数据属于 Evidence;一次实验结论进入 Experience/报告;多次都稳定成立的实验设计原则才进入 Method。可重复的采样脚本可进入 benchmark/Skill,但不要把语义裁决写进脚本。

来源:`docs/local/METHOD-PLAYBOOK-v0.1.md` §7;最近 working-set receipts Qwen A/B 为主要实证样本。

From deyi2026/llm-first-loop, Apache-2.0.

More from deyi2026/llm-first-loop

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Sep 8, 2026 (1 month ago)
Last release
Oct 22, 2026 (8 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from deyi2026

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get method-ab-experiment running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying method-ab-experiment?

CallGet help
Cite this page

method-ab-experiment skill. Agent Mart. https://agentmart.dev/skills/deyi2026-method-ab-experiment (last updated 2026-10-24)