系統SYS-2026-0104v0.1
XA-06L——本地真實模型執行交接包
從 READY_FOR_REAL_MODEL 到密封結果包的 Windows/本地交接:configure → preflight → run/resume → verify/analyze → seal。resume 只在試驗結果與 XA-03 摘要都在、兩個 SHA-256 仍相符、身分相符且 XA-03 狀態為 COMPLETE 時才跳過該點;被改過的點變成 INVALID,絕不靜默覆寫。封存排除機密值、以來源紀錄遮蔽絕對路徑、把執行目錄改為相對於封存包並全部重新雜湊,讓封存包能在另一台機器上重新驗證。軟體 29/29 PASS,套件內的實驗狀態:NOT_RUN——「READY_FOR_REAL_MODEL ≠ REAL_MODEL_PILOT_COMPLETE」正是套件要說的重點。
目的
purpose- Let a local machine with a GPU execute the canonical 36-trial pilot against an OpenAI-compatible server and hand back a relocatable, hash-sealed bundle — including an honest REAL_MODEL_PILOT_INCOMPLETE when a gate fails.
架構
architecture- thin PowerShell wrappers over a cross-platform core: config, preflight, matrix, runner, state (pilot_state.json with per-point hashes), gate, seal; IMPORT_BACK.md defines what to upload for analysis.
倉庫與站點
documentation- README.md, SPEC.md, IMPLEMENTATION_PLAN.md, IMPORT_BACK.md, STATUS.json, validation_report.json inside the package
記錄欄位
tags- EML-IPM-XA-06L
- v0.1
- FINAL_RELEASE_BUILT_AFTER_CLEAN_CANDIDATE_GATE
關係
| 來源 | 關係 | 目標 | 狀態 | ID |
|---|---|---|---|---|
SYS-2026-0104 XA-06L——本地真實模型執行交接包 | extends | SYS-2026-0103 XA-06——真實模型 pilot 閘 | ACTIVE | REL-2026-0458 |
SYS-2026-0104 XA-06L——本地真實模型執行交接包 | supports | RES-2026-0103 能力線——拿掉 LOOP 還剩多少智能,以及統一的智能事件 | ACTIVE | REL-2026-0459 |
SYS-2026-0104 XA-06L——本地真實模型執行交接包 | released_as | ART-2026-0118 EML-IPM-XA-06L v0.1 — local real-model execution handoff pack (configure → preflight → run → seal) artifact://evemisslab/intelligence-physical-metrology/IPM_v0.2_XA06L_Local_Execution_Handoff_v0.1.zip | ACTIVE | REL-2026-0460 |
EXP-2026-0103 XA-06 第一次真實模型 pilot——Qwythos-9B-v2 走 A0→A5 階梯(36 試驗,2026-09-03) | runs_on | SYS-2026-0104 XA-06L——本地真實模型執行交接包 | ACTIVE | REL-2026-0478 |
歷史與來源歷程
- Canonical URL
- https://evemisslab.com/ai/systems/SYS-2026-0104/
- 快照
AI-SNAPSHOT-v0.1-fe85b9694a45- 來源歷程
source- EveMissLab research collection: Intelligence Physical Metrology (真本體論13)
extracted_by- Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports
extracted_at- 2026-09-11
generator- tools/extract_all.py
claim_boundary- status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports