Files
MeltMind/docs/o1-v2-panel-stop-results-2026-11-17.md
T

156 lines
11 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# O1 v2:同60件预检未达到共同几何覆盖门
2026-11-17 05:35:31 Asia/Shanghai,唯一队列已停止,实际协调541230及全部O1 worker消失。原始终态为 V2_PANEL_IDENTIFIABILITY_STOP,global_stop=null、monitor_errors=[]、accounting_complete=true。systemd保留真实ExecMainPID541230/exit1及开始/结束时间;exit1是预定科学门停止,不是基础设施故障。无重试、参数放宽、第三版设计、后续几何或拟合。
|预定检查|实际结果|门槛|结论|
|---|---:|---:|---|
|件级共同几何合格 E/60|14/60 = 23.3333%|≥80%|未过|
|合格件两接口同时达标 H/E|14/14 = 100%|严格>70%|通过|
|全60中同时合格且达标 H/60|14/60 = 23.3333%|描述项|—|
|面板组件完整配对|2/21 = 9.5238%|≥80%|未过|
|配对组件两接口均值达标|2/2 = 100%|严格>70%|通过|
21是此非随机调度面板实际出现的组件数,不是全库209。组件只有面板内所有出现件均合格才配对,先分别平均两接口再等权计数;同组件未入面板成员完整列于panel-feasibility.json。原60包含v1全部11不合格、3中止,无替换。不能把14/14当全库成功率,也未运行完整209组件门。
## 资格失败与干预分开解释
60个工程guard全部PASS,科学结果14PASS+46INELIGIBLE,每件恰好G/T/C三次seed0调用,共180次,无候选挑选。所有60都有三臂接口差异。12件至少一臂实际体积平衡不合格:G4件、T3件、C6件,其中1件G/C同时失败。其余48件三臂单独物理条件通过,其中34件三臂互检不合格。互检失败维度可重叠:15件排序体积分数配对差越界、22件切面面积max/min>1.25、2件排序直径max/min>1.25;不是39个不同件。48件可计算三臂互检中,原±5%口径通过0件;其余12件的三臂互检未计算,已因单臂物理条件失败而不合格。此项仅描述,不替换v2判据。
在14件共同合格子集,TG和TC接口差均≥0.10,说明本次类型面权确能形成接口干预;覆盖不足使预定实验无法在此共同域继续。v2同时改生成目标与形状界,不能用v1/v2差分别归因两项。接口改变不等于预测改善;没有训练或预测收益结果。结论限定为此数据、K8、固定类型/读出与一次v2干预未达到预定可辨识条件,不否定所有类型、拓扑或真实连通路线。
## 固定来源、审计与费用
主审Claude1794863409/h-2e3bd8ed;完整Mold规格SHA632688217e59dd134922245e524ff1c0f63933be64600be21379f7aa74947a5f。实现a946adfb0222998be9b6d68776239a0afc26c164,合同90caf09948fe2b0cc51c54c7cf87986f5daaa86f222dd2d1cf231bf5bc70a874,原60名单c4193a57dc2fa942ffe3e87f7f23c885f2caa5aa83509cf7ecb110e700b935d3,门结果5b82bf8a13b353533db3f575080396e35e6f764d2da457296efb06734e0b774a。源码/合同/名单独立远端HTTP哈希读回PASS。14项合成通过;终态只读审计逐件回执、输出指纹与全部552数组SHA通过,source pins PASS,不重新解码、分区、拟合或改原记录。
新60 worker369.598083CPU秒;含旧账/新合成/metadata预留的队列累计1001.411572CPU秒,加独立只读审计guard2.960984,最终累计1004.372556/43200CPU秒。guard75.679929091、协调34.662186425另列。科学队列墙钟134.481641秒。原v1全部60记录及625.377883费用一次保留,原失败不被v2替换。
551其余可准备件从未尝试,6原缺特征仍保留;全库617/209身份不变。没有标签解码、模型、预测或评分。按主审停止此尺度,不进入全库准备或O1拟合,不请求重复go。交Claude/Mold按既有阶段规则统一回顾;R4bc独立原队列不因本结果改变。
原始文件根artifacts/o1-v2-preparation-2026-11-17:panel-parts.csv/json含每件状态、原组件、三接口、体积/面积/直径、成本及指纹;panel-feasibility.json含全部组件和未入面板同组件成员;parts含全部图、面权、置换、分区、读出、日志和失败;panel-independent-audit.json、terminal-process-audit.json、final-cost-ledger.json保存独立核验。小型审阅证据与完整原始数组按顺序发布,远端完成状态以发布回执为准。
R4bc05:36只读快照:原521868在,626终态=625PASS+原job90,fit78795.511600CPU秒;无新失败/globalstop/monitorerror,batch未终态。没有改动其代码、队列或预算。
## 60件表
|件号|科学状态|原因|TG|TC|GC|
|---|---|---|---:|---:|---:|
|3|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.959389|0.969157|0.950021|
|4|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.938932|0.956475|0.890804|
|6|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.986131|0.963119|0.994069|
|8|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.977439|0.981929|0.970897|
|9|PASS|—|0.981939|0.985839|0.974126|
|11|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.982817|0.978935|0.961219|
|21|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.988602|0.979286|0.976240|
|23|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.953148|0.960676|0.966580|
|24|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.937219|0.956768|0.941372|
|30|PASS|—|0.972604|0.988063|0.970652|
|31|PASS|—|0.971742|0.979080|0.972552|
|49|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.940245|0.950702|0.895279|
|57|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.975084|0.964305|0.961720|
|58|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.976378|0.980848|0.971634|
|60|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.976667|0.985516|0.975863|
|61|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.969663|0.980243|0.969943|
|65|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.939638|0.923456|0.933062|
|66|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.973352|0.976801|0.986344|
|67|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.968228|0.973679|0.985155|
|68|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.945616|0.944527|0.932326|
|70|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.886098|0.940046|0.842099|
|72|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.894072|0.953269|0.910502|
|77|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.969441|0.976787|0.977935|
|78|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.942246|0.953537|0.912041|
|79|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.994816|0.993057|0.983050|
|87|PASS|—|0.945725|0.919835|0.904625|
|88|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.993851|0.966514|0.970901|
|93|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.996086|0.985216|0.983268|
|98|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.979501|0.989453|0.984031|
|102|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.981619|0.972397|0.973374|
|103|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.927603|0.921479|0.843601|
|104|PASS|—|0.978527|0.980992|0.992298|
|105|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.966062|0.969933|0.969651|
|106|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.960982|0.989365|0.950500|
|111|PASS|—|0.958079|0.979769|0.995533|
|115|PASS|—|0.981213|0.983383|0.978343|
|116|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.965578|0.967122|0.936732|
|117|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.962536|0.961935|0.936314|
|119|PASS|—|0.966558|0.975987|0.977738|
|120|PASS|—|0.911760|0.959311|0.931078|
|125|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.860562|0.939012|0.920277|
|128|PASS|—|0.989857|0.972685|0.990063|
|129|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.995791|0.979968|0.991720|
|130|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.991578|0.991644|0.977001|
|131|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|1.000000|0.988998|0.992075|
|132|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.999260|0.996466|1.000000|
|133|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.999809|0.997062|0.979793|
|140|PASS|—|0.973255|0.973742|0.965248|
|142|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.975759|0.993503|0.969304|
|144|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.949126|0.962119|0.953607|
|145|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.980273|0.978753|0.959429|
|152|PASS|—|0.985087|0.988102|0.972071|
|153|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.973625|0.992558|0.990132|
|154|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.955661|0.962046|0.947215|
|157|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.968662|0.967113|0.925952|
|161|PASS|—|0.974258|0.961065|0.959636|
|162|PASS|—|0.947053|0.965062|0.965527|
|163|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.973711|0.954484|0.957142|
|164|INELIGIBLE|V2_TRIPLE_MATCH_FAIL|0.965808|0.951004|0.939523|
|165|INELIGIBLE|V2_ARM_GEOMETRY_FAIL|0.962156|0.941868|0.938477|
## 逐件约束失败分解(原结果只读派生)
12件单臂物理体积失败之后,三臂互检未执行;这些件的其他互检维度是未检验,不记通过。34件互检失败可在多维同时越界,下表不去重删除失败。
|件号|实际失败约束|
|---|---|
|3|T: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|4|cut area max/min >1.25|
|6|cut area max/min >1.25|
|8|pair volume >0.025;cut area max/min >1.25|
|11|pair volume >0.025|
|21|cut area max/min >1.25|
|23|cut area max/min >1.25|
|24|pair volume >0.025|
|49|cut area max/min >1.25|
|57|cut area max/min >1.25|
|58|pair volume >0.025|
|60|pair volume >0.025|
|61|pair volume >0.025|
|65|cut area max/min >1.25|
|66|C: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|67|pair volume >0.025|
|68|G: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|70|pair volume >0.025|
|72|cut area max/min >1.25|
|77|sorted diameter max/min >1.25|
|78|C: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|79|sorted diameter max/min >1.25|
|88|cut area max/min >1.25|
|93|pair volume >0.025|
|98|cut area max/min >1.25|
|102|pair volume >0.025|
|103|G: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|105|G: GeometryFailure: PHYSICAL_VOLUME_BALANCE; C: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|106|T: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|116|cut area max/min >1.25|
|117|cut area max/min >1.25|
|125|pair volume >0.025|
|129|pair volume >0.025;cut area max/min >1.25|
|130|cut area max/min >1.25|
|131|cut area max/min >1.25|
|132|pair volume >0.025;cut area max/min >1.25|
|133|cut area max/min >1.25|
|142|C: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|144|pair volume >0.025;cut area max/min >1.25|
|145|T: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|153|cut area max/min >1.25|
|154|G: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|157|pair volume >0.025;cut area max/min >1.25|
|163|C: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
|164|cut area max/min >1.25|
|165|C: GeometryFailure: PHYSICAL_VOLUME_BALANCE|
Claude1794864979.171309已接受按预裁量停止,下一步仅本逐件分解/最终审计及R4bc原队列全池/INTERIM_100精确比对,再由Claude统一写阶段收束与建议。本报告将结论限定为规定构造与匹配门在此面板不足,不推断其他所有K8构造不可能。
## 完整原始证据发布完成
原单次push59077退出0,完整552数组/1,457,411,422B固定e7fa46da9b393c49a5ab29e2d040c30f1d9253e5。远端ref与完整文件清单核验通过;首件G分区、末件C分区与审计文件独立HTTP/SHA核验通过,不声称完整1.46GB二次下载。详见raw-publication-complete.json。此前小型审阅入口53d95182已获Claude1794865219接受;科学与逐件结果不变,旧±5%口径明确为48件可计算互检/12件未计算。Claude关于面积匹配可能约束干预的设计启示保留为推断与未来建议,未据此开启新试验。