{
  "comparisons": {
    "q3-revised-vs-baseline": {
      "relative_change_pct": -0.6927149477129024,
      "paired_bootstrap_95_pct": [
        -4.498553709698716,
        2.946109407935289
      ],
      "faster_cases": 9,
      "runs": 18
    },
    "q3-revised-vs-legacy": {
      "relative_change_pct": 1.6296939233583885,
      "paired_bootstrap_95_pct": [
        -1.639291360386112,
        5.191944866303212
      ],
      "faster_cases": 8,
      "runs": 18
    },
    "q4-revised-vs-baseline": {
      "relative_change_pct": -14.378947944816856,
      "paired_bootstrap_95_pct": [
        -17.92759937325484,
        -11.480044833864715
      ],
      "faster_cases": 18,
      "runs": 18
    },
    "q4-revised-vs-legacy": {
      "relative_change_pct": -2.261375942162047,
      "paired_bootstrap_95_pct": [
        -6.079501282327849,
        1.7095590628835715
      ],
      "faster_cases": 13,
      "runs": 18
    }
  },
  "official": [
    {
      "passed": true,
      "source": "official_practice",
      "strategy": "rl",
      "problem": 3,
      "case_code": "F38Y-9Q9J-B82V-CB9Z",
      "official_jammer_count": 13,
      "directional_jammer_count": 0,
      "cleared_channels": [
        2,
        3,
        4,
        5,
        6,
        8,
        9,
        10,
        11,
        14,
        15,
        19,
        20
      ],
      "accepted_requests_including_enter_exit": 126,
      "measures": 111,
      "clear_attempts": 13,
      "clear_successes": 13,
      "virtual_time_s": 3538.811466,
      "real_elapsed_server_s": 2.321,
      "checks": {
        "official_problem_matches_summary": true,
        "official_package_sha256_matches": true,
        "normal_enter_and_exit": true,
        "responses_within_official_run_window": true,
        "unique_accepted_request_ids": true,
        "all_official_sources_cleared": true,
        "summary_completed_and_consistent": true,
        "all_timestamps_match_microsecond_recomputation": true,
        "within_real_and_virtual_limits": true
      },
      "limits": "仅验证本官方演练案例；不触碰正式测试，不解读官方加密日志内容。"
    },
    {
      "passed": true,
      "source": "official_practice",
      "strategy": "rl_q4",
      "problem": 4,
      "case_code": "X43X-VMRF-8GRZ-3SYA",
      "official_jammer_count": 10,
      "directional_jammer_count": 3,
      "cleared_channels": [
        2,
        7,
        10,
        12,
        15,
        16,
        17,
        18,
        19,
        20
      ],
      "accepted_requests_including_enter_exit": 621,
      "measures": 585,
      "clear_attempts": 34,
      "clear_successes": 10,
      "virtual_time_s": 14003.086251,
      "real_elapsed_server_s": 5.723,
      "checks": {
        "official_problem_matches_summary": true,
        "official_package_sha256_matches": true,
        "normal_enter_and_exit": true,
        "responses_within_official_run_window": true,
        "unique_accepted_request_ids": true,
        "all_official_sources_cleared": true,
        "summary_completed_and_consistent": true,
        "all_timestamps_match_microsecond_recomputation": true,
        "within_real_and_virtual_limits": true
      },
      "limits": "仅验证本官方演练案例；不触碰正式测试，不解读官方加密日志内容。"
    },
    {
      "passed": true,
      "source": "official_practice",
      "strategy": "rl_q4",
      "problem": 4,
      "case_code": "BXSA-TFQ3-2DJ6-DHH7",
      "official_jammer_count": 10,
      "directional_jammer_count": 10,
      "cleared_channels": [
        2,
        4,
        5,
        6,
        8,
        11,
        12,
        13,
        14,
        17
      ],
      "accepted_requests_including_enter_exit": 693,
      "measures": 670,
      "clear_attempts": 21,
      "clear_successes": 10,
      "virtual_time_s": 14562.81673,
      "real_elapsed_server_s": 6.39,
      "checks": {
        "official_problem_matches_summary": true,
        "official_package_sha256_matches": true,
        "normal_enter_and_exit": true,
        "responses_within_official_run_window": true,
        "unique_accepted_request_ids": true,
        "all_official_sources_cleared": true,
        "summary_completed_and_consistent": true,
        "all_timestamps_match_microsecond_recomputation": true,
        "within_real_and_virtual_limits": true
      },
      "limits": "仅验证本官方演练案例；不触碰正式测试，不解读官方加密日志内容。"
    }
  ],
  "feature_probe": {
    "mismatched_indices": [
      7,
      14,
      15,
      17,
      18
    ],
    "fixed_history_count": 0.0833333358168602,
    "old_history_count": 0.0
  },
  "bc_probe": {
    "q3": {
      "samples": 1029,
      "different_argmax_fraction": 0.909621000289917,
      "max_logit_difference": 10.808266639709473
    },
    "q4": {
      "samples": 1013,
      "different_argmax_fraction": 0.674234926700592,
      "max_logit_difference": 32.212066650390625
    }
  },
  "limitations": "置信区间仅重采样固定模型的场景，不覆盖训练随机种子不确定性。",
  "code_sha256": {
    "src\\mcm26b\\rl\\__init__.py": "670d3134aec149e9a2e03cde00cb56c3ab16c590584ff0afd1f522212ebd2242",
    "src\\mcm26b\\rl\\_constants.py": "88998fee55f227c755e8770a6e8aca0abd2508ab417696b0c1f7e45c189f74eb",
    "src\\mcm26b\\rl\\env.py": "d01a11a5ebb5eff921b095442caee05b007a8d61a1aea154831cd1e24f46bf7b",
    "src\\mcm26b\\rl\\features.py": "406995743e0f7fbe09581f7aca55cf28882149c4b7f2eae2aba89ef32c69fdcf",
    "src\\mcm26b\\rl\\policy.py": "5bee6329f14c5b5e88173e63624e68292949db5b7b92d3dcabf3948f8aa53755"
  }
}