第10章 复现实验流程(训练/推理/对齐)


I. 目标与范围(Purpose & Scope)


II. 输入与依赖(Prerequisites & Inputs)


III. 流程总览(E2E Flow)


IV. 关键对齐(时间→路径→相位)


V. 执行细则(Executable Steps)


VI. 产物与日志(Artifacts & Logs)


VII. 放行判定(Release Decision)


VIII. 质量门映射(Gates Mapping)


IX. 机读配置(Machine-Readable Configs)
A. eval/compare_spec.yaml(节选)

version: "1.0.0"

coverage: { mode: "k", k: 2 } # k|alpha|quantile

metrics:

mae: { tolerance: 1.0e-4 }

auc: { tolerance: 2.0e-3 }

r_phi: { lb95_min: 0.60 }

delta_t_arr_s: { guard: "tau_T_s" }

epsilon_flux_p95: { guard: 0.02 }

latency_p95_s: { guard: 0.200 }

rules:

interval_overlap_required: true

same_coverage_band_required: true


B. tools/compare.py 输出(示例)

{

"decision": "pass",

"deltas": { "MAE": 1.0e-5, "Latency_P95_s": 0.006 },

"intervals": { "r_phi_ref":[0.61,0.80], "r_phi_repro":[0.62,0.79], "overlap": true },

"gates": { "G1": true, "G2": 0.94, "G3": true, "G4": true, "G5": true, "G6": true, "G7": true, "G8": true }

}


X. 反例与修正(Anti-Patterns & Fixes)


XI. 交叉引用(Cross-References)


XII. 勾选清单(Checklist)