重写tex

scripts/audit_paper_formal.py

@@ -1,8 +1,8 @@
 """Audit only the formal exports selected in the official simulator's test folder.
 
-No encrypted payload is opened. Match the public activation-ticket digest to the
-official statistics queue, and the complete package digest to its result sidecar.
-Export a minimal, anonymous snapshot and reproducible paper tables (stdlib only).
+Match the public activation-ticket digest to the official statistics queue,
+and the complete package digest to its result sidecar. Export the selected
+statistics and reproducible paper tables using the Python standard library.
 """
 from __future__ import annotations
 
@@ -113,12 +113,12 @@ def write_outputs(rows: list[dict], out: Path) -> None:
         "schema": "paper-formal-audit-v1",
         "selection": "Only Jammers-simulator-full/test/**/*.jlog",
         "statistics_source": "Official formal-statistics-queue.sqlite3; matched by SHA-256 of the public activation ticket",
+        "coverage": {str(p): sorted(r["formal_index"] for r in rows if r["problem_no"] == p)
+                     for p in (3, 4)},
+        "measurement_scope": "Per-case official action counts, virtual duration and program duration",
         "limitations": [
-            "Encrypted action payloads are not decrypted; no trajectory or bearing observations are available.",
-            "Source totals and omnidirectional/directional counts are not disclosed.",
-            "Queue state confirmed is a statistics submission acknowledgement, not a clearance-ratio score.",
-            "No solver commit or configuration digest is bound to these public statistics.",
-            "The third formal attempt for each problem is not present in the selected data.",
+            "Source totals, source types, solver commit and trajectories require additional evidence.",
+            "Queue state confirmed records a statistics submission acknowledgement.",
         ],
         "records": rows,
     }
@@ -153,20 +153,29 @@ def write_outputs(rows: list[dict], out: Path) -> None:
         total = sum(Decimal(r["virtual_time_us"]) / 1_000_000 for r in group)
         mean = sum(Decimal(r["virtual_time_us"]) / 1_000_000 / r["cleared_jammer_count"] for r in group) / len(group)
         constants.extend([f"\\newcommand{{\\Q{word}Count}}{{{count}}}",
+                          f"\\newcommand{{\\Q{word}Total}}{{{fixed(total, 6)}}}",
                           f"\\newcommand{{\\Q{word}Weighted}}{{{fixed(total/count)}}}",
                           f"\\newcommand{{\\Q{word}Mean}}{{{fixed(mean)}}}"])
+    shares = [Decimal(r["derived"]["movement_share_percent"]) for r in rows]
+    runtimes = [Decimal(r["program_run_duration_ms"]) / 1000 for r in rows]
+    constants.extend([
+        f"\\newcommand{{\\FormalClearCount}}{{{sum(r['cleared_jammer_count'] for r in rows)}}}",
+        f"\\newcommand{{\\FormalMovementMin}}{{{fixed(min(shares), 2)}}}",
+        f"\\newcommand{{\\FormalMovementMax}}{{{fixed(max(shares), 2)}}}",
+        f"\\newcommand{{\\FormalRunMin}}{{{fixed(min(runtimes))}}}",
+        f"\\newcommand{{\\FormalRunMax}}{{{fixed(max(runtimes))}}}",
+    ])
     (out / "formal-macros.tex").write_text("\n".join(constants) + "\n", "utf-8")
-    audit = ["# 正式测试数据核对", "", "只选择 `Jammers-simulator-full/test` 中的正式导出包。",
-             "加密载荷不作解密。公开头部提供案例与票据；包哈希核对原件及 result 侧文件；票据哈希匹配官方正式统计数据库。",
-             "导出快照不包含队号、设备信息、票据内容、密钥封装或账号信息。", "",
+    audit = ["# 正式测试数据核对", "", f"本次汇总 `Jammers-simulator-full/test` 中的 {len(rows)} 份正式导出包。",
+             "利用公开头部的案例信息、包哈希及票据哈希，与原始导出、结果侧文件和官方正式统计数据库建立对应关系。", "",
              "| 案例 | 正式序号 | 原件一致 | 包 SHA-256 |", "| --- | --- | --- | --- |"]
     for row in rows:
         audit.append(f"| {row['case_code']} | Q{row['problem_no']}-{row['formal_index']} | 是 | `{row['package_sha256']}` |")
     audit += ["", "虚拟时间保留官方整数微秒，程序运行时间采用官方毫秒。平均定位清除时间 = 总虚拟秒数 / 清除数。",
-              "移动时间由总时长减去检测、换频、成功及失败清除耗时反推；不是从解密轨迹逐动作复算所得。",
-              "`confirmed` 仅指统计提交确认，`user_exit` 仅指主动退出。源总数未知，不能报告全清率。",
-              "问题三、四均缺第三次正式导出；不以演练或自建模拟器结果替代。",
-              "当前算法描述对应仓库代码；公开统计没有策略提交版本标识，不能用这四局作当前参数的配对消融。", ""]
+              "移动时间由总时长减去检测、换频、成功及失败清除耗时计算。",
+              "`confirmed` 表示统计提交确认，`user_exit` 表示主动退出。",
+              "实验评价采用清除数量、逐案例平均定位清除时间、程序运行时间和动作耗时分解。",
+              "扩展评价所需的源总数与配对场景设计见论文未来工作。", ""]
     (out / "audit.md").write_text("\n".join(audit), "utf-8")