u0 3 threshold stability route - yasufumi-nakata/mind-upload GitHub Wiki

Wiki: U0-3 threshold-stability route packet

A row-specific packet for the threshold-governance identity row

This learning page is generated for GitHub Wiki. The public portal is managed on mind-upload.com.

Updated: 2026-04-04 / Role: U0-3 route packet

Role Of This Page

This page fixes the current public route for U0-3 at the level of one question, one intended EEG route anchor, one bounded first artifact, and one funding bridge.

Accuracy Notes

This page does not claim personhood continuity or branch-identity resolution. It only records what the present EEG route can honestly close first.

Back To Public Pages

Related Wiki Pages

Wiki: Current public six route brief - The compact six-row public route.
Wiki: Deep focus notes - The wider six-row working note.
Wiki: Grant and dataset playbook - The fixed funding and anchor map.

U0-3 threshold-stability route packet

This page does not change the current public six-question batch. It makes only the fourth active row more explicit, so readers can see why U0-3 stays next after U0-2, why D02 CSTE now stays only as a gated support anchor, and what the next public-safe artifact can honestly close.

What still stays fixed after the 14:10 JST actual-pull reread on April 4, 2026

An actual-pull recheck at 2026-04-04 14:10 JST after mind-upload@cafeab4bc2a5, EEG-DATA@e0695bcac83d, and auto-research-funds@a4852d89376c still keeps U0-3 fourth and still keeps D02 only as the planned gated-support anchor for the present threshold-governance question.
The newest pulled deltas remain carry-forward rather than route replacement. mind-upload tightens glossary excitability and ionic terms on top of the earlier EEG foundation-model source-status split without widening the route, EEG-DATA still reports 30,930 merged-unique rows, 28,893 high-confidence rows, 24,909 immediate rows, and 24,539 high-confidence immediate rows after the newest metadata deep dives, and auto-research-funds adds JELA Diakonia scholarship cards without changing the practical shortlist.
None of those additions replaces the intended threshold floor. The newest carry-forwards still do not supply the cross-session / cross-task human-biometrics framing that keeps D02 primary in planning, while ds004148 and ds007216 still stay the main public-open threshold stress-test ladders.
Browser-level live checks in this run still returned HTTP/2 200 for the current public entry pages and this row packet with Last-Modified: Sat, 04 Apr 2026 04:18:04 GMT; the checked front-door pages already surfaced 2026-04-04 12:02 JST wording while this row packet still surfaced 2026-04-04 09:05 JST wording before this source sync. A fresh read-only Todoist audit still showed the existing funding-task coverage as sufficient (items=206, projects=2, Kura=2, Nakatani family=11, Kashinome=2, Secom=3, Brain Science Foundation=8, Okawa=2, Japan Industrial Science Research Institute=1, Doumei=0, Kenkyu no Susume=0), so no new Todoist item was justified for this row in this sync.

What still stays fixed after the 09:05 JST actual-pull reread on April 4, 2026

An actual-pull recheck at 2026-04-04 09:05 JST after mind-upload@e571567f017d, EEG-DATA@fbbb0d8c01e0, and auto-research-funds@d7f6fb9bc158 still keeps U0-3 fourth and still keeps D02 only as the planned gated-support anchor for the present threshold-governance question.
The newest pulled deltas remain carry-forward rather than route replacement. mind-upload now just carries the earlier 08:07 JST route-hold sync on top of a recent pretraining-overlap audit, EEG-DATA still reports 30,929 merged-unique rows, 28,892 high-confidence rows, 24,909 immediate rows, and 24,539 high-confidence immediate rows after the OpenNeuro ds007020 deep dive, and auto-research-funds adds deeply reviewed April 4 scholarship updates without changing the practical shortlist.
None of those additions replaces the intended threshold floor. The newest carry-forwards still do not supply the cross-session / cross-task human-biometrics framing that keeps D02 primary in planning, while ds004148 and ds007216 still stay the main public-open threshold stress-test ladders.
Browser-level live checks in this run still returned HTTP/2 200 for the current public entry pages and this row packet with Last-Modified: Sat, 04 Apr 2026 00:11:41 GMT; the checked front-door pages already surfaced 2026-04-04 08:07 JST wording while this row packet still surfaced 2026-04-04 01:04 JST wording before this source sync. A fresh read-only Todoist audit still showed the existing funding-task coverage as sufficient (items=206, projects=2), so no new Todoist item was justified for this row in this sync.

What still stays fixed after the 01:04 JST actual-pull reread on April 4, 2026

An actual-pull recheck at 2026-04-04 01:04 JST after mind-upload@9b6a8b971ba0, EEG-DATA@56d765da60b, and auto-research-funds@3c8569e28 still keeps U0-3 fourth and still keeps D02 only as the planned gated-support anchor for the present threshold-governance question.
The newest pulled deltas remain carry-forward rather than route replacement. mind-upload now carries the earlier 00:06 JST route reread plus field-level artifact-stack disclosure tightening, EEG-DATA still reports 30,921 merged-unique rows, 28,884 high-confidence rows, 24,900 immediate rows, and 24,530 high-confidence immediate rows after deepening Numbers and PRIOS, and auto-research-funds adds current coverage for 同盟育成会 and 研究のすゝめ奨学金 without changing the practical shortlist.
None of those additions replaces the intended threshold floor. The newest carry-forwards still do not supply the cross-session / cross-task human-biometrics framing that keeps D02 primary in planning, while ds004148 and ds007216 still stay the main public-open threshold stress-test ladders.
Browser-level live checks in this run still returned HTTP/2 200 for the current public entry pages and this row packet with Last-Modified: Fri, 03 Apr 2026 21:11:06 GMT; the checked row packet still surfaced Updated: 2026-04-02 before this source sync even though the front-door pages already surfaced the 2026-04-04 01:04 JST basis. A fresh read-only Todoist audit still showed the existing funding-task coverage as sufficient (items=206, projects=2, Kura=2, Nakatani family=11, Nakatani graduate scholarship=7, Kashinome=2, Secom=3, Brain Science Foundation=8, Okawa=2, Japan Industrial Science Research Institute=1, Keio graduate scholarship=2, G-7=1, ENEOS=1, 同盟会=0, 研究のすゝめ=0), so no new Todoist item was justified for this row in this sync.

What still stays fixed after the 23:08 JST actual-pull reread on April 2, 2026

An actual-pull recheck at 2026-04-02 23:08 JST after mind-upload@f5d1d1dd1683, EEG-DATA@ee1463d28288, and auto-research-funds@ec24a355abbc still keeps U0-3 fourth and still keeps D02 only as the planned gated-support anchor for the present threshold-governance question.
The newest pulled deltas remain carry-forward rather than route replacement. mind-upload clarifies the causal-versus-human-proxy bridge outside the current six, EEG-DATA now reports 30,896 merged-unique rows, 28,859 high-confidence rows, and 24,882 directly usable rows after candidate-curation refinement plus the OpenNeuro ds007615 deep dive, and auto-research-funds adds Geo-Kagaku current cards without changing the practical shortlist.
None of those additions replaces the intended threshold floor. The latest public-open carry-forwards still do not supply the cross-session / cross-task human biometrics framing that keeps D02 primary in planning, and ds004148 plus ds007216 still stay the main public-open stress-test ladders.
Browser-level live checks still returned HTTP/2 200 for the current public entry pages with Last-Modified: Thu, 02 Apr 2026 13:12:21 GMT, and a fresh read-only Todoist audit still showed the existing funding-task coverage as sufficient, so this page only refreshes the row-specific route-hold wording.

What still stays fixed after the 12:04 JST actual-pull reread on April 2, 2026

An actual-pull recheck at 2026-04-02 12:04 JST after mind-upload@dceb8bc, EEG-DATA@04b7fda167e, and auto-research-funds@4442e348a still keeps U0-3 fourth and still keeps D02 only as the planned gated-support anchor for the present threshold-governance question.
The newest pulled deltas remain carry-forward rather than route replacement. mind-upload syncs a spectroscopy front door outside the current six, EEG-DATA now reports 30,888 merged-unique rows, 28,851 high-confidence rows, and 24,506 immediate rows after THINGS-EEG, an OSF orientation route, and a rat ECoG deep dive, and auto-research-funds only refreshes repository stats.
None of those additions replaces the intended threshold floor. PhysioNet auditory-eeg is public-open but still too short-horizon and too low-channel to replace the cross-session / cross-task identity-drift floor, while THINGS-EEG and rat ECoG remain the wrong task family for the present threshold-stability packet. ds004148 and ds007216 still stay the main public-open stress-test ladders.
Browser-level live checks still returned 200 for the current public entry pages with Last-Modified: Thu, 02 Apr 2026 02:37:04 GMT, and a fresh read-only Todoist audit still showed the existing funding-task coverage as sufficient, so this page only refreshes the row-specific route-hold wording.

What was rechecked for this row

D02 CSTE still fixes the intended repeated-session, cross-session / cross-task identity-drift floor that matches U0-2, but the current IEEE DataPort page is subscription-gated, so it can no longer be described here as an immediate public route.
The current funding bridge remains narrow and unchanged: Nakatani Foundation graduate scholarship stays first and Kashinome Scholarship stays second as a runway-support route.
The public six-question batch also stays unchanged: U13-2 / U13-5 / U0-2 / U0-3 / U14-4 / U14-2.

Why `U0-3` still stays fourth

It is the threshold-governance row that only becomes interpretable after U0-2 has already fixed the timing floor.
If this row fails, stronger threshold-stability wording should stay blocked because the route would still be unable to separate bounded identity drift from split-specific overfit.
The row can keep threshold selection narrow and benchmark-like before any broader continuity language is attempted.

Why `D02` stays in scope only as a gated support anchor

Candidate	What it helps with	Why it is not the first public anchor
`D02 CSTE` (`10.21227/j162-nh61`, subscription-gated support)	It still fixes the intended repeated-session and cross-task floor shared with `U0-2`, which keeps the `threshold stability + overfit exclusion` question in scope if access is available.	The current IEEE DataPort page is subscription-gated, so it no longer counts as an immediate public anchor in this run.
`A test-retest resting and cognitive state EEG dataset` (`10.18112/openneuro.ds004148.v1.0.1`, immediate)	It gives a richer session-external retest ladder that may later stress-test how stable the threshold rule remains outside the present biometric framing.	It is useful as a retest adjunct, but it is not the cleanest direct identity-drift floor for the first public threshold packet.
`A multi-session simultaneous EEG-fMRI dataset with online experience sampling` (`10.18112/openneuro.ds007216.v1.0.0`, immediate)	It gives denser multi-session state labels and richer logging that may later support threshold-stability stress tests under stronger state drift.	It is multimodal and experience-sampling oriented rather than the narrowest current EEG-only threshold floor, so it stays a support anchor.

Among public-open backups, ds004148 still stays the cleanest retest ladder and ds007216 still stays the cleanest multi-session provenance ladder. They are strong threshold stress-test families, but they still do not replace the intended identity-drift floor that keeps D02 primary in planning.

What the first artifact can honestly close now

Artifact piece	What it closes now	What still stays outside the claim
`Threshold-stability note`	A bounded report of how much the decision threshold moves across tasks and sessions on the same identity-drift floor as `U0-2`.	Any statement that threshold stability by itself settles personhood continuity or branch identity.
`Overfit-exclusion rule`	A narrow rule for rejecting models that only look stable under split-specific or single-session conditions.	Any claim that one accepted threshold already proves a broader continuity benchmark.
`Selection-rationale table`	A public-safe record of why one threshold survives cross-session reading better than nearby alternatives.	Any theory-level conclusion that goes beyond the present EEG artifact.

First-pass KPI bundle and stop rule

Item	Current row-level fixation
`First-pass KPI bundle`	`session-out AUC`, `threshold stability`, `overfit-exclusion rate`
`Stop rule`	Do not widen this row if the threshold order flips across sessions or if accepted models remain split-specific after the overfit screen is applied.

Funding wording that still fits

Route	Submission wording that stays honest	Why it still fits this row
`Nakatani Foundation graduate scholarship`	`Doctoral-track threshold-stability and overfit-exclusion benchmark for identity drift in the mind-upload route`	The row still fits a doctoral-track identity-drift benchmark when kept narrow and audit-first.
`Kashinome Scholarship`	`Master's-stage threshold-governance rerun pack for identity drift in the mind-upload route`	It still fits only as a runway-support layer rather than the main long-horizon route.

External dependency tasks that stay outside this session

A session-external rerun pack is still needed before this row can support stronger threshold wording.
A larger cohort extension is still needed before the route can claim broader threshold robustness.
Any personhood-continuity or branch-identity interpretation remains outside the present EEG-only route.

What changed in this run

The public batch did not change.
The main change in this sync is that U0-3 now carries the same 2026-04-04 09:05 JST route basis as the latest front-door reading and records that the live row packet had still been stuck at the earlier 2026-04-04 01:04 JST wording before this source update.
The row still records the anchor-choice rationale, one compact first-pass KPI bundle, and one explicit stop rule instead of leaving them implicit.
This is still a route-setting update, not a solved-claim update.