메뉴
当你在为《自动化学报》(Acta Automatica Sinica, AAS) 设计与呈现实验、仿真或实物验证时调用。覆盖研究问题(RQ)驱动的实验设计、公平基线与对照、控制类的稳定性/收敛性/鲁棒性分析、模式识别类的指标与统计显著性、消融实验、仿真与实物平台的报告要素，帮助为控制理论、机器人、智能系统方向的中文稿搭建经得起 Acta Automatica Sinica 双盲外审的实证与理论证据链。
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Use when preparing the accountability artifacts that accompany an ACM FAccT paper — datasheets for datasets, model cards, data statements, audit and impact-assessment documentation, and released code/data — since FAccT's culture centers documentation and accountability infrastructure rather than a formal ACM artifact-badging track; covers what makes each genre credible, anonymized-review versus public-release versions, and consistency with the paper's harm claims.
Use when drafting ACM FAccT author responses — the short factual-correction rebuttal after preliminary reviews, and, distinctively for the 2026-new process, the Revise-and-resubmit round where you address Area-Chair-prioritized concerns from a mixed CS+law+social-science panel, map each to a concrete change, and survive a re-review while keeping mutual anonymity.
Use when preparing an accepted ACM FAccT paper for camera-ready — de-anonymizing and adding the withheld Positionality/Acknowledgements statements, switching to the acmart sigconf proceedings format, ACM metadata (DOI, ORCID, CCS concepts, rights), the two-round camera-ready calendar for Accept vs Revise papers, ACM Open Access, finalizing datasheets/model cards, and integrating required changes without scope creep.
Use when designing or auditing ACM FAccT empirical work — quantitative fairness audits with disaggregated metrics and fair baselines, qualitative and participatory studies with coding and reflexivity, mixed-methods designs, sound handling of protected attributes and proxies, consent and IRB for human-subjects and community-facing work, and matching evidence to the shape of a fairness/accountability/transparency claim.
Use when positioning an ACM FAccT submission across its many disciplinary lanes — algorithmic fairness/ML, HCI, law and policy, STS and critical theory, and prior FAccT/FAT* proceedings — writing delta-first contrast that a mixed reviewer pool will accept, citing borrowed constructs to their real origin, keeping self-citations mutually anonymous, and declaring overlap with workshops, preprints, and prior versions.
Use when strengthening ACM FAccT transparency and reproducibility — releasing code, data, and analysis for quantitative audits; documenting datasets and models with datasheets, model cards, and data statements; making qualitative and participatory work auditable without breaking confidentiality; pinning provenance for scraped and model-generated data; and keeping the paper, the supplementary material, and any released artifact consistent.
| name | aas-experiments |
| description | 当你在为《自动化学报》(Acta Automatica Sinica, AAS) 设计与呈现实验、仿真或实物验证时调用。覆盖研究问题(RQ)驱动的实验设计、公平基线与对照、控制类的稳定性/收敛性/鲁棒性分析、模式识别类的指标与统计显著性、消融实验、仿真与实物平台的报告要素，帮助为控制理论、机器人、智能系统方向的中文稿搭建经得起 Acta Automatica Sinica 双盲外审的实证与理论证据链。 |
本技能帮你为《自动化学报》(Acta Automatica Sinica, AAS) 搭建经得起双盲外审的证据链。控制/自动化
稿件的证据不只是"跑得好"，还包括理论保证（稳定性、收敛性）与公平对照。以下方法于
2026-07-09 核验框架（见 resources/official-source-map.md）。
先把要回答的研究问题写清，再据此设计实验：
RQ1 所提控制律是否保证闭环稳定/收敛？——理论证明 + 仿真验证
RQ2 相比基线，性能指标是否显著改进？——对比实验 + 定量指标
RQ3 各设计模块是否必要？——消融实验
RQ4 对扰动/参数摄动是否鲁棒？——鲁棒性/敏感性实验
| 要点 | 做法 |
|---|---|
| 基线选择 | 选最接近、最强的已有方法，非弱化版稻草人 |
| 同等条件 | 相同系统模型、初始条件、噪声、算力预算 |
| 调参公平 | 基线也认真调参，报告调参协议 |
| 多场景 | 多个被控对象/数据集，避免单点结论 |
aas-reproducibility）。| 形式 | 用途 |
|---|---|
| 响应曲线/相图 | 展示动态过程与稳定性 |
| 对比表 | 多方法多指标定量对齐 |
| 消融表 | 各模块贡献 |
| 箱线/误差棒 | 多次运行的分布与方差 |
图表遵循 aas-writing-style：图下表上、矢量、黑白可辨、法定单位。
[RQ] 每个实验对应明确 RQ？
[理论] 稳定性/收敛性有严格证明与假设？
[基线] 最强公平基线？同等条件？调参协议已报告？
[指标] 口径统一、任务匹配？
[统计] 多次运行均值±方差？必要显著性检验？
[消融] 关键模块逐一验证？
[鲁棒] 扰动/摄动/噪声扫描 + 失效边界？
[复现] 环境/种子/数据固定（见 aas-reproducibility）？
[图表] 符合本刊体例？
aas-writing-style 修正。不同方向的实验骨架略有差异，可据此起步：
| 方向 | 典型实验骨架 |
|---|---|
| 自适应/鲁棒控制 | 稳定性证明 → 仿真跟踪/调节 → 参数摄动鲁棒 → 与经典控制对比 |
| 多智能体一致性 | 拓扑设定 → 收敛性证明 → 不同拓扑仿真 → 通信/饱和约束验证 |
| 强化学习控制 | 环境设定 → 多种子训练曲线 → 与基线策略对比 → 泛化/鲁棒 |
| 模式识别 | 数据集划分 → 主指标对比 → 消融 → 混淆矩阵/可视化 |
| 机器人 | 仿真验证 → 实物平台 → 多工况 → 演示（可作附件） |
[归因] 改进能归因到具体设计而非调参运气？（消融支撑）
[稳健] 换数据集/对象/种子结论是否稳定？
[对照] 与最强基线在同等条件下比较？
[统计] 差异是否显著、是否报告方差？
[复现] 配置/种子/环境是否可复现（见 aas-reproducibility）？
《自动化学报》(Acta Automatica Sinica, AAS) 的证据期待是"理论分析 + 充分实验"双支撑：控制类
强调稳定性/收敛性证明与仿真/实物验证，识别/智能类强调统计严谨与公平对照。让实验设计服务于坐实
可数贡献，而非堆砌图表，配合 aas-reproducibility 让 Acta Automatica Sinica 的外审对结果可靠性
放心。