现状：到 2026-10-21 11:31 EDT 为止Where things stand (as of 2026-10-21 11:31 EDT)
数据截至：2026-10-21 11:31 EDT（cognition-os/08_RUNNING.md 第 9 版）。信任级别用 cognition-os/03_TRUST_TABLE.md 的状态词。Data as of: 2026-10-21 11:31 EDT (cognition-os/08_RUNNING.md, revision 9). Trust states use the words of cognition-os/03_TRUST_TABLE.md.
1. 已确立（Trusted 或 Provisional）1. Established (Trusted or Provisional)
- NGC 5746 的中心超额不是仪器在探测器位置上制造的。MOS \(+0.98 \pm 0.18\)（5.5\sigma）、pn \(+0.87 \pm 0.14\)（6.4\sigma），都是科学值减去它自己两个零场（10.1–30 kpc）。The NGC 5746 central excess is not made by the instrument at that detector position. MOS \(+0.98 \pm 0.18\) (5.5\sigma), pn \(+0.87 \pm 0.14\) (6.4\sigma), both science minus its own two nulls (10.1–30 kpc).
- pn 的零信号偏置为零。六个 pn 零场在 10.1–30 kpc 的反方差平均 \(-0.02 \pm 0.06\)。The pn null-signal offset is zero. Inverse-variance mean of six pn nulls at 10.1–30 kpc: \(-0.02 \pm 0.06\).
- MOS 的零信号偏置不为零，且随场而变。九个 MOS 零场平均 \(-0.19 \pm 0.06\)；七个常规场 \(-0.37 \pm 0.08\)，两个硬超额场 \(+0.06 \pm 0.10\)（两组相差 \(0.43 \pm 0.12\)，3.4\sigma）。The MOS null-signal offset is not zero and is field-dependent. Nine MOS nulls mean \(-0.19 \pm 0.06\); seven normal fields \(-0.37 \pm 0.08\), two hard-excess fields \(+0.06 \pm 0.10\) (the groups differ by \(0.43 \pm 0.12\), 3.4\sigma).
- 两条已拒绝的解释。冻结软质子谱指数不是原因（SP-free \(-0.44 \pm 0.13\) vs 冻结 \(-0.43 \pm 0.11\)，R003）；pn QPB 通过联合天空泄漏进 MOS 也不是原因（MOS-only \(-0.35 \pm 0.11\) vs 联合 \(-0.43 \pm 0.11\)，D004 在阈值处拒绝，R004）。Two explanations rejected. The frozen soft-proton index is not the cause (SP-free \(-0.44 \pm 0.13\) vs frozen \(-0.43 \pm 0.11\), R003); nor is pn QPB leaking into MOS through the joint sky (MOS-only \(-0.35 \pm 0.11\) vs joint \(-0.43 \pm 0.11\), rejected at the D003 threshold, R004).
- 稳定天空把常规场偏置移到 \(-0.27 \pm 0.08\)（之前 \(-0.37 \pm 0.08\)），但按 D005 预注册规则这是"不确定"（D006）；引数规则：MOS 科学值与零值必须来自同一拟合变体。The stable sky moves the normal-field offset to \(-0.27 \pm 0.08\) (was \(-0.37 \pm 0.08\)), but by the pre-registered D005 rule the verdict is ambiguous (D006); quoting rule: a MOS science value and its null must come from the same fit variant.
- 现实 Chandra 中央星系选样（Provisional）。用真实指向、转角、活动芯片与群表选样后，Zhang 红移范围（z 0.02–0.10）内 50/100/150/200/250 kpc 处分别有 1507/1276/1069/913/769 颗中央星系，约 60 Ms 唯一曝光。最初的覆盖普查（Chandra \(\approx 3.7 \times 10^{3}\) 到 150 kpc）被红队复核判 Questioned：三条规则被推翻（指向坐标而非目标中心、ACIS-S 真实半径 1.59′ 而非 4.15′、XMM 可去星系团场），中央星系选样以 I009 为准。The realistic Chandra central sample (Provisional). With real pointings, roll, active chips and the group catalogue, the Zhang redshift range (z 0.02–0.10) gives 1507/1276/1069/913/769 centrals at 50/100/150/200/250 kpc, about 60 Ms of unique exposure. The original coverage census (Chandra \(\approx 3.7 \times 10^{3}\) to 150 kpc) is Questioned after a red-team review: three rules were overturned (pointing rather than target centre, the true ACIS-S radius 1.59′ instead of 4.15′, XMM cluster fields can be removed); the central selection follows I009.
- Zhang 归一化质疑已被复核为稳健。数字化 Zhang 图 5 的 10–30 kpc 点是 \(\beta\) 模型中位数的 \(1.576\times\) 倍；缩放后 K/pred 原始 \(0.107 \pm 0.063\)、加 \(+0.3\) K 后 \(0.209 \pm 0.063\)——矛盾仍小于 1 倍，但 10–20 kpc 没有独立的 Zhang 点。The Zhang-normalisation objection was re-checked and is robust. The digitised Zhang Fig. 5 point at 10–30 kpc is \(1.576\times\) the \(\beta\)-model median; scaled K/pred is \(0.107 \pm 0.063\) raw and \(0.209 \pm 0.063\) with \(+0.3\) K — the tension stays under a factor of 1, but there is no independent Zhang point at 10–20 kpc.
来源：cognition-os/insights/I006（横幅与 (e),(k),(m)）；cognition-os/insights/I007、I009；cognition-os/07_REVISION_LOG.md R001–R011；cognition-os/04_DECISION_LOG.md D003–D006。Source: cognition-os/insights/I006 (banner and (e),(k),(m)); cognition-os/insights/I007, I009; cognition-os/07_REVISION_LOG.md R001–R011; cognition-os/04_DECISION_LOG.md D003–D006.
2. 仍开放2. Open
- MOS 零场偏置的原因未定。三个候选都没找到驱动：天空范型（(a) 没有天空范数预测 18 个场侧的 A）、硬波段 QPB/软质子（(b)）、MOS1 CCD 组（(c)）。但天空在必须稳定的地方不稳定：同一场的 N/S 两侧给出不同天空（LHB 差 100 倍），四个 NGC 5746 历元给出 CXB 0 到 \(3.8 \times 10^{-7}\)。The cause of the MOS null offset is open. Three candidates found no driver: the sky model ((a): no sky norm predicts A over the 18 field-sides), the hard-band QPB/soft protons ((b)), and the MOS1 CCD set ((c)). But the sky is not stable where it must be: the N and S reference boxes of one field give different skies (LHB differs by 100\times), and the four NGC 5746 epochs give CXB from 0 to \(3.8 \times 10^{-7}\).
- 哪一步处理造成 \approx 5 倍矛盾：六条候选，一条都没检验。（盘光计入 CGM；Zhang 内区是去卷积模型；平均 vs 中位数；点源深度；我们的系统误差 kT/背景/CCD1；背景定义与自减。）Which processing step makes the factor \(\approx 5\): six candidates, none tested. (Disk light counted as CGM; Zhang's inner profile is a deconvolved model; mean vs median; point-source depth; our own systematics kT/background/CCD1; background definition and self-subtraction.)
- XMM 档案样本能否用：剔除 AGN 与星暴目标后，15–50 Mpc 只剩约 15–20 个干净 L* 旋涡（约 1 Ms）。Is the XMM archival sample usable: after removing AGN and starburst targets only about 15–20 clean L* spirals at 15–50 Mpc (\(\approx 1\) Ms) remain.
- Vrot vs M*：无直接数据；"Vrot 更好"的论断（0.18 vs 0.33 dex）被红队质疑——所用 SMHM 斜率 \(\approx 0.49\) 在银河系质量处低估，本地斜率 \(\approx 1.11\) 时 M* 代理 \(\approx 0.14\) dex 反而更紧（RT-05 复核中）。eRASS1 分组检验本身仍成立。Vrot vs M*: no direct data; the "Vrot is better" argument (0.18 vs 0.33 dex) is questioned by a red team — the SMHM slope \(\approx 0.49\) used is low at MW mass, and with the local slope \(\approx 1.11\) the M* proxy is \(\approx 0.14\) dex, tighter (RT-05 recomputing). The eRASS1 split test itself still stands.
- Chandra 银河系质量档 50–150 kpc 到不了 \(3\sigma\)：即使只算光子统计，0.5–1.2 keV 的 S/N 在 50–100 kpc 为 2.32、100–150 kpc 为 0.97（当前设计 no-go）；现实试点里同口径银河系质量 10–25 kpc 也只有 S/N \(\approx 1.6\)。转向高质量（log M* 11–11.5）10–50 kpc 正控（N=107，1.68 Ms，预报 S/N \(\approx 5\)）——它是流程检验，不检验银河系质量矛盾。The MW-mass Chandra stack at 50–150 kpc is not reachable at \(3\sigma\): even photon-only, the 0.5–1.2 keV S/N is 2.32 at 50–100 kpc and 0.97 at 100–150 kpc (the current design is a no-go); in the realistic pilot the same-geometry MW-mass 10–25 kpc gives only S/N \(\approx 1.6\). The plan redirects to a high-mass (log M* 11–11.5) 10–50 kpc positive control (N=107, 1.68 Ms, forecast S/N \(\approx 5\)) — a tool test, not a test of the MW-mass tension.
- eRASS1 Vrot 分组预报：等权平均表面亮度估计量下 \(b = 0.6/0.88/1.3\) 给 \(10.2/14.1/23.4\sigma\)（未标定 ECF、未计内禀散布，条件化统计预算）；预注册判定规则（\(\Delta V \ge 0.30\) dex 且 \(Z \ge 3\) 为正；90% CI 在 \(\pm 0.15\) dex 内为零）已获 PI 批准，在任何 X 射线数据叠加前冻结。The eRASS1 Vrot split forecast: under the equal-weight mean surface-brightness estimator, \(b = 0.6/0.88/1.3\) gives \(10.2/14.1/23.4\sigma\) (uncalibrated ECF, intrinsic scatter not included — a conditional statistical budget); the pre-registered decision rule (\(\Delta V \ge 0.30\) dex and \(Z \ge 3\) for a positive; 90% CI inside \(\pm 0.15\) dex for a null) is PI-approved and frozen before any X-ray data are stacked.
来源：cognition-os/insights/I006 (l)、I010；cognition-os/insights/I001、I003；cognition-os/02_CURRENT_QUESTION.md；07_REVISION_LOG.md R012–R017；04_DECISION_LOG.md D010、D011；P12/P13/P14 报告。Source: cognition-os/insights/I006 (l), I010; cognition-os/insights/I001, I003; cognition-os/02_CURRENT_QUESTION.md; 07_REVISION_LOG.md R012–R017; 04_DECISION_LOG.md D010, D011; the P12/P13/P14 reports.
3. 信任表（T01–T11）3. Trust table (T01–T11)
| 编号ID | 条目Item | 状态Status | 信任Trust |
|---|---|---|---|
| T01 | ADAP \(\ge 2\) 代理共识表（1015 星系 / 82.6 Ms，2018 截止）ADAP \(\ge 2\)-agent consensus list (1015 galaxies / 82.6 Ms, 2018 cutoff) | Provisional | 中medium |
| T02 | 分辨叠加 \(0.17 \pm 0.10\)（10–20 kpc，Vc 标度，4 星系）；工作区间 \(0.25\)–\(0.37 \pm 0.10\)（MOS 零值偏置、背景自减、新 NGC 2613 提取后）Resolved combination \(0.17 \pm 0.10\) (10–20 kpc, Vc scaling, 4 galaxies); working range \(0.25\)–\(0.37 \pm 0.10\) after the MOS null offset, background self-subtraction and the newer NGC 2613 extraction | Questioned | 低–中（Zhang 归一化后验未传播；RT-02/RT-02b）low–medium (Zhang normalisation posterior not propagated; RT-02/RT-02b) |
| T03 | 矛盾原因清单the tension-causes list | Questioned | 不适用（均未检验）n/a (none tested) |
| T04 | eROSITA 方法在 XMM 指向档案上的假设违反分析the analysis of eROSITA-method assumptions violated by the XMM pointed archive | Provisional | 中medium |
| T05 | kpc-在-场内 设计窗口与自减比例kpc-in-field design windows and self-subtraction ratios | Trusted | 高（作为算术）high (as arithmetic) |
| T06 | 零信号对照叠加设计null-signal control stack design | Provisional | 中medium |
| T07 | NGC 5746 超额也在 pn 中（不主要是 MOS CCD1）NGC 5746 excess also in pn (not mainly MOS CCD1) | Provisional | 中medium |
| T08 | NGC 5746 超额非仪器所致；MOS 零值不为零（随场变化，原因未定）NGC 5746 excess not instrumental; MOS nulls NOT zero (field-dependent, cause open) | Provisional | 中medium |
| T09 | 覆盖普查（Chandra \(3.7 \times 10^{3}\) SDSS MW 质量；Vrot 选样只有几十）coverage census (Chandra \(3.7 \times 10^{3}\) SDSS MW-mass; Vrot-selected only tens) | Questioned | 远样本中等、近样本低（RT-01 三条规则被推翻；中央星系选样另见 I009）medium for distant samples, low for nearby (RT-01 overturned three rules; central selection superseded by I009) |
| T10 | eRASS 与 Chandra 极限之间的星表级点源光：原始面亮度是 10–50 kpc 叠加 CGM 的 2–4 倍（未扣 logN–logS 背景负项）catalogue-level point-source light between the eRASS and Chandra limits: raw surface brightness 2–4 times the stacked 10–50 kpc CGM (the logN–logS background negative term is not closed) | Provisional | 低low |
| T11 | Vrot 是更好的晕代理：scatter(log M200 | Vrot) \(\approx 0.18\) dex vs scatter(log M200 | M*) \(\approx 0.33\) dex（均为反演）Vrot is the better halo proxy: scatter(log M200 | Vrot) \(\approx 0.18\) dex vs scatter(log M200 | M*) \(\approx 0.33\) dex (both inversions) | Questioned | 中（RT-04 质疑所用 SMHM 斜率；RT-05 复核中）medium (RT-04 questions the SMHM slope used; RT-05 recomputing) |
来源：cognition-os/03_TRUST_TABLE.md（逐行照抄，2026-10-21 07:59 EDT 版）。Source: cognition-os/03_TRUST_TABLE.md (copied row by row, the 2026-10-21 07:59 EDT version).
4. 待定：仍在运行的任务4. Pending: the tasks still running
以下任务在运行，结果在报告被分诊前一律待定。本手册不预测它们的结果。已分诊的任务见第 2 节与"本次更新"。The following tasks are running; their results are pending until the reports are triaged. This handbook does not predict them. Triaged tasks appear in section 2 and in "What's new".
| 编号ID | 任务Task | 产出步骤Producing step |
|---|---|---|
| P6 | Chandra 试点 20 个 ObsID 的还原（Virgo）Chandra pilot reduction of 20 ObsIDs (Virgo) | 实测背景底限裁决 S/N 预报the measured background floor decides the S/N forecast |
| P8 | eRASS1 He & Li 复现（Virgo）eRASS1 He & Li reproduction (Virgo) | P12 预注册检验的流水线the pipeline for the P12 pre-registered test |
| P15 | eRASS1 热谱计数转换（HEASARC 镜像）eRASS1 thermal count conversion (HEASARC mirror) | 预报的能段/谱形转换the band/spectral conversion the forecast needs |
| P16 | Chandra 零检验设计（含高质量试点的晕质量与速度弥散）Chandra null-test design (halo mass and velocity dispersion of the high-mass pilot) | D011 试片的对照设计the control design for the D011 pilot |
| RT-05 | 红队：SMHM 斜率（攻击 RT-04 的结论）red team: the SMHM slope (attacking the RT-04 conclusion) | "Vrot 更好"论断的斜率基础the slope basis of the "Vrot is better" argument |
| RT-06 | 红队：Chandra S/B 重算red team: the Chandra S/B recompute | S/N 预报的独立复核independent re-analysis of the S/N forecast |
来源：cognition-os/RESEARCH_PLAN.md；cognition-os/SEARCH_QUEUE.md；cognition-os/REDTEAM_QUEUE.md；cognition-os/08_RUNNING.md §3。Source: cognition-os/RESEARCH_PLAN.md; cognition-os/SEARCH_QUEUE.md; cognition-os/REDTEAM_QUEUE.md; cognition-os/08_RUNNING.md §3.
5. 本次更新（2026-10-21 11:31 EDT）5. What's new (2026-10-21 11:31 EDT)
- Chandra 银河系质量档 50–150 kpc 在当前设计下到不了 \(3\sigma\)。即使只算光子统计（0.5–1.2 keV），50–100 kpc 的 S/N 为 2.32、100–150 kpc 为 0.97；原"完全相关 5% 背景项"的论证被推翻（同场差分后该项不成立）。当前设计 no-go；P14 寻找可达设计。The MW-mass Chandra stack at 50–150 kpc is a no-go in the current design. Even photon-only (0.5–1.2 keV), the S/N is 2.32 at 50–100 kpc and 0.97 at 100–150 kpc; the original "fully correlated 5% background term" argument is overturned (it does not survive same-field differencing). The current design is a no-go; P14 looks for reachable designs.
- Chandra 试点转向高质量 10–50 kpc 正控（D011）。log M* 11–11.5、z 0.005–0.20、ACIS-I+S、2010 年前：N=107、111 个 ObsID、1.68 Ms，预报 S/N \(\approx 5\)（\(\sigma_c = 2\%\) 时仍 4.06）。它是流程/正控，不是银河系质量矛盾的检验（同口径 MW 10–25 kpc 只有 S/N \(\approx 1.6\)）。The Chandra pilot redirects to a high-mass 10–50 kpc positive control (D011). log M* 11–11.5, z 0.005–0.20, ACIS-I+S, pre-2010: N=107, 111 ObsIDs, 1.68 Ms, forecast S/N \(\approx 5\) (still 4.06 at \(\sigma_c = 2\%\)). It is a tool/positive control, not a test of the MW-mass tension (the same-geometry MW 10–25 kpc gives only S/N \(\approx 1.6\)).
- eRASS1 Vrot 分组预报改为 \(10.2/14.1/23.4\sigma\)（\(b = 0.6/0.88/1.3\)）。原 \(39\sigma\) 用的是计数加权；等权平均表面亮度估计量下为 \(23.4\sigma\)（\(b = 1.3\)）。仍未标定 ECF、未计内禀散布。预注册规则已获 PI 批准（D010），在任何 X 射线数据叠加前冻结。The eRASS1 Vrot split forecast becomes \(10.2/14.1/23.4\sigma\) (\(b = 0.6/0.88/1.3\)). The original \(39\sigma\) used count weighting; the equal-weight mean surface-brightness estimator gives \(23.4\sigma\) at \(b = 1.3\). The ECF is still uncalibrated and intrinsic scatter is not included. The pre-registered rule is PI-approved (D010) and frozen before any X-ray data are stacked.
- "Vrot 是更好代理"的论断被质疑（RT-04）。0.33 dex 用了 SMHM 斜率 \(\approx 0.49\)；银河系质量处本地斜率 \(\approx 1.11\) 时 M* 代理 \(\approx 0.14\) dex，反而比 Vrot 更紧。RT-05 复核中。Vrot 分组检验本身（子问题 B）仍成立。The "Vrot is the better proxy" argument is questioned (RT-04). The 0.33 dex used an SMHM slope \(\approx 0.49\); with the local slope \(\approx 1.11\) at MW mass the M* proxy is \(\approx 0.14\) dex, tighter than Vrot. RT-05 is recomputing. The Vrot split test itself (sub-question B) still stands.
- Zhang 归一化质疑复核为稳健（P9）。数字化 Zhang 图 5 的 10–30 kpc 点是 \(\beta\) 模型中位数的 \(1.576\times\) 倍；缩放后 K/pred 原始 \(0.107 \pm 0.063\)、加 \(+0.3\) K 后 \(0.209 \pm 0.063\)。矛盾仍小于 1 倍，但 10–20 kpc 没有独立的 Zhang 点。The Zhang-normalisation objection re-checked as robust (P9). The digitised Zhang Fig. 5 point at 10–30 kpc is \(1.576\times\) the \(\beta\)-model median; scaled K/pred is \(0.107 \pm 0.063\) raw and \(0.209 \pm 0.063\) with \(+0.3\) K. The tension stays under a factor of 1, but there is no independent Zhang point at 10–20 kpc.
- 现实 Chandra 中央星系选样完成（P10/P3）。Zhang 红移范围内 50/100/150/200/250 kpc 处 1507/1276/1069/913/769 颗；西半天 50 kpc 只有 398 颗。母体不是 Zhang 的精确 30825（质量标尺中位差 \(-0.12\) dex）。The realistic Chandra central selection is done (P10/P3). 1507/1276/1069/913/769 at 50/100/150/200/250 kpc in the Zhang redshift range; the western half-sky has only 398 at 50 kpc. The parent is not Zhang's exact 30825 (mass-scale median offset \(-0.12\) dex).
数据截至Data as of ：2026-10-21 11:31 EDT（cognition-os/08_RUNNING.md 第 9 版）。: 2026-10-21 11:31 EDT (cognition-os/08_RUNNING.md, revision 9).