Wiki

Mind-upload RQ60 RQ-by-RQ Deep Dossiers (EEG-DATA × auto-research-funds)

汎用要約ではなく、RQ単位で実装可能な深掘り

Mind Uploading Research Project

Public Page Updated: 2026-04-04 RQ-by-RQ Deep Dossier

How to use this page

Read this first to avoid getting lost

Best for
60RQを個別理解して応募計画へ落とし込みたい研究者
Reading time
55-80分
What the wiki is for

The wiki is a learning aid. For the project's official current synthesis, success criteria, and operating rules, always return to the public pages.

Mind-upload RQ60 RQ-by-RQ Deep Dossiers (EEG-DATA × auto-research-funds)

本ページは、60RQ を1件ずつ深掘りするための固定ドシエ集です。
汎用横断サマリは最小化し、各RQで 検証命題 境界 応募テーマ 主データ KPI 停止条件 を先に固定します。

2026-04-04 22:06 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@39248d59caaa, EEG-DATA@7ba2201e2b5, and auto-research-funds@f90279417a9f still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-04 08:07 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@418c8f5d0ef9e871531d9cfb8548887e57f54d68, EEG-DATA@dc6d833b83e9a5e6a181edfee5349f561f65b043, and auto-research-funds@58d8738be5d55cff13d7bd626479783d6818de1f still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-03 05:12 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@a9fc5d1cb691f3669c40803604e78cf87fab4eed, EEG-DATA@8c3ace28bad0b3bac259caa440994e5d05bfc7d1, and auto-research-funds@35756f8381c47daa24bb70bb7a1ff88d1dcf76bc still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-03 04:06 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@f5e13ebdc4748ce683d905c37cabfdb3c53036bb, EEG-DATA@53f5c8166faf0d5fcebfb5c39741d4992eb82a39, and auto-research-funds@1f190fec93a9180a27443b1fe2b31f9488dfbf74 still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-02 18:06 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@85b6603b569b, EEG-DATA@a7f9390492fd, and auto-research-funds@b6c0d0f4eeb7 still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-02 11:09 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@208e59e2180c, EEG-DATA@00d85380f3f, and auto-research-funds@f4ba35518400 still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-02 10:07 JST latest actual-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@be507c810b11, EEG-DATA@3f20f48cca6e, and auto-research-funds@fb32c05163f4 still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-01 20:09 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@da9644095ba6, EEG-DATA@1ecc5cdc0131, and auto-research-funds@b5cd37febbbb still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-01 17:06 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@c5a0550, EEG-DATA@82893f6ca6e, and auto-research-funds@ec6f7d068 still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-01 14:08 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. An actual-pull reread after mind-upload@ccef40f, EEG-DATA@deefd5f95c6, and auto-research-funds@f5f76c485 still leaves the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-01 12:06 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@9f4bc8e760b1, EEG-DATA@1324f9e1ae2b, and auto-research-funds@6548b1650535 still leave the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-01 11:06 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@aa21ab9855c1, EEG-DATA@cccd137a3468, and auto-research-funds@3a691196cb9f still leave the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-04-01 03:09 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@31835fcd8daa, EEG-DATA@12eef8ca4ef0, and auto-research-funds@040e5b8798fd still leave the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-03-31 21:09 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@ed6f8203e46d, EEG-DATA@20fbbad9959f, and auto-research-funds@77f189662745 still leave the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-03-31 19:07 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@b7b6c0498ae1, EEG-DATA@24e85cc20f09, and auto-research-funds@0fddd53bd03c still leave the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-03-31 17:06 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@4731680627ff, EEG-DATA@3e7a53d9ca8, and auto-research-funds@ef2924014a65 still leave the current D10 / D03 / D02 / D01 family and the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1 unchanged.

2026-03-31 16:05 JST latest post-pull route-hold note

This run still does not change the current public six-question batch. The latest pulled deltas were reread before writing this note: mind-upload@a11240d38415 tightened the connectome / timing-state route cards, EEG-DATA@f318ac8ce42d added an OSF eucqf deep-dive refresh, and auto-research-funds@c95cc61fdfc9 added new Takeda / Nohken funding-card material. None of those changes replace the current D10 / D03 / D02 / D01 anchor family or the outside-current-six order U7-1 -> U1-2 -> U4-2 -> U8-1.

2026-03-31 00:04 JST overnight route-hold note

This run still does not change the current public six-question batch. The added value is not a new batch switch but a stronger reason for the hold: the latest EEG-DATA upstream additions remain outside the current speech / identity / reproducibility anchor family, D10 remains the only clearly public-open primary anchor, and D03/D02/D01 now have to be read only as subscription-gated support anchors rather than immediate public anchors.

2026-03-30 live evidence note for the current public six

This run keeps the current public batch unchanged. The new contribution is row-level evidence wording that can be quoted safely in public pages without inflating the claim ceiling. Each active row now has one explicit EEG-DATA reason, one bounded first artifact, and one funding sentence that stays consistent with the current shortlist.

RQ EEG-DATA reason that still justifies the row First artifact that can be closed now Funding sentence that still fits No-go boundary that stays explicit
U13-2 D10 3M-CPSEED remains the strongest current speech anchor because the local mirror still exposes an immediate OpenNeuro route with overt, silent-intended, and imagined speech across 20 participants x 4 sessions, plus raw BIDS and phase-split derivatives. Brain-minus-prior control table and mimic-separation note. Keep Kura Fund first and Nakatani Foundation graduate scholarship second, because the row still reads naturally as a bounded young-researcher / doctoral-track speech-decoding benchmark rather than a solved mind-upload claim. Do not translate this row into evidence of preserved internal mechanism, preserved identity, or successful mind preservation.
U13-5 D03 Multimodal Fusion EEG + Eye-Tracking remains only a gated planning anchor because the intended paired EEG and gaze route is still useful, but the visible DataPort metadata centers a 55-student cognitive-load dataset and does not expose a recall-specific label on the public page. Perception-to-recall degradation note and transfer-gap table. Keep Kura Fund first and Nakatani Foundation graduate scholarship second, because the row is still a decoder-transfer benchmark with a clear failure mode rather than a broader mechanism claim. Do not write this as evidence that shared decoder structure implies shared internal mechanism across perception and recall.
U0-2 D02 CSTE remains the cleanest intended identity-drift floor because the local mirror still places it in Biometric / Authentication and Cross-session / Transfer / Domain adaptation, with 10 participants, even though the current DataPort page is subscription-gated. Offset/jitter distribution and state-feature collapse audit. Keep Nakatani Foundation graduate scholarship first and Kashinome Scholarship second, because the row still fits a student-facing continuity / drift audit better than a broad systems proposal. Do not elevate timing stability or re-identification robustness into legal, philosophical, or normative identity continuity.
U0-3 The same D02 anchor remains correct because threshold stability only becomes interpretable after the identity row stays on a cross-session / cross-task biometric floor instead of a single-session score. Threshold-stability note and overfit-exclusion rule. Keep Nakatani Foundation graduate scholarship first and Kashinome Scholarship second, because the row still reads as a doctoral-track identity-drift benchmark with bounded operational claims. Do not present a stable threshold as a sufficient condition for personhood continuity or branch-identity resolution.
U14-4 D01 Acquisition delay of wireless EEG remains acceptable only as a bounded entry point, because the visible package is a latency/synchronization dataset with README_en plus device/run .mat files rather than a rerun benchmark by itself. Filled Model Card / Dataset Card examples for the current batch. Keep Kura Fund first and Nakatani Foundation research grant as next-cycle watch, because the row is still strongest as a reproducibility/disclosure package rather than a current open BME result claim. Do not turn card completeness into field-wide reproducibility, deployment readiness, or standards adoption.
U14-2 The same D01 anchor remains enough to freeze the split boundary only as a latency-first floor; broader split or rerun wording still needs D13 as the first public-open support family. Fixed-split rerun contract and negative-case trail. Keep Kura Fund first and Nakatani Foundation research grant as next-cycle watch, because the row still functions as a bounded reproducibility-contract artifact. Do not present exploration outputs as confirmatory evidence or imply broader reproducibility than the current anchor can support.

2026-03-30 13:08 JST row packets for the current public six

The current public batch still does not change. What changes in this addendum is the working granularity for the six active rows: each one now has one submission-ready title, one explicit anchor reason, one bounded first artifact, and one sentence about what still blocks stronger wording.

RQ Submission-ready title Current first artifact Why the present anchor is enough now Funding bridge that still fits What still blocks stronger wording
U13-2 Neural-contribution audit for imagined speech with a brain-minus-prior control Brain-minus-prior control table and mimic-separation note D10 3M-CPSEED still keeps overt, silent-intended, and imagined speech inside one public release, so the first artifact can compare a neural contribution against a prior-only shortcut instead of collapsing into a generic decoder score. Kura Fund, then Nakatani Foundation graduate scholarship The row still does not preserve internal mechanism, preserved identity, or successful mind preservation.
U13-5 Perception-to-recall transfer-gap benchmark for mimic separation Perception-to-recall degradation note and transfer-gap table D03 Multimodal Fusion EEG + Eye-Tracking still provides the intended paired EEG + gaze route, but the visible DataPort metadata is still a cognitive-load package and does not expose a recall-specific label on the public page, so it only freezes a bounded planning branch point rather than an immediate public anchor. Kura Fund, then Nakatani Foundation graduate scholarship The row still cannot support a stronger preserved-mechanism reading without an external recall-validity layer and a richer public-open adjunct benchmark.
U0-2 Temporal-validity-aware identity drift audit with offset, jitter, and state-feature collapse Offset/jitter distribution and state-feature collapse audit D02 CSTE remains cross-session, cross-task, and biometric, but the current DataPort page is subscription-gated, so identity-like wording can first be reduced only to a bounded timing-and-feature audit. Nakatani Foundation graduate scholarship, then Kashinome Scholarship The row still cannot cross the line into legal, philosophical, or normative identity continuity.
U0-3 Task-specific threshold stability and overfit-exclusion benchmark for identity drift Threshold-stability note and overfit-exclusion rule The same D02 anchor keeps threshold movement on the same cross-session / cross-task floor as U0-2, which is the minimum condition for interpreting threshold stability at all. Nakatani Foundation graduate scholarship, then Kashinome Scholarship The row still cannot treat a stable threshold as sufficient for personhood continuity or branch-identity resolution.
U14-4 Card-integrated reproducibility contract for the current public batch Filled Model Card / Dataset Card examples D01 Acquisition delay of wireless EEG is narrow, but it is still enough to force latency and measurement fields into concrete card examples that a rerun contract can actually reference; stronger public wording would still need D13 as the first public-open rerun family. Kura Fund; Nakatani Foundation research grant as next-cycle watch The row still cannot justify field-wide reproducibility, deployment readiness, or standards uptake wording.
U14-2 Fixed-split rerun contract for exploration versus confirmation Fixed-split rerun contract and negative-case trail The same D01 route is still sufficient to make an exploration / confirmation split operational as a narrow latency floor, but stronger rerun wording would still need D13 as the first public-open rerun family. Kura Fund; Nakatani Foundation research grant as next-cycle watch The row still cannot support broad rerun-robustness wording or confirmatory language for exploratory outputs.

今回の再検証ログ（2026-03-30 02:05 JST, one-RQ closure gates）

本runでは、current public 6RQ の batch 自体は変えずに、「今この6件で何を閉じたら public-safe に進捗と呼べるか」を RQ ごとに明文化いたしました。ここで固定するのは solved claim ではなく、1RQ=1最初の成果物=1非主張境界 の closure gate です。

RQ 今回の closure gate この gate を先に閉じる理由 主アンカー claim ceiling を上げる前に次に要るもの
U13-2 brain-minus-prior control table と mimic-separation note language decoder を一括りにせず、neural contribution の最小監査を先に閉じる必要があるため。 D10 external language-evaluation layer と control-condition audit
U13-5 perception-to-recall degradation note と transfer-gap table branch point を先に固定しないと mimicry / mechanism の過読が残るため。 D03 recall-label validity と adjunct speech/control benchmark
U0-2 offset/jitter distribution と state-feature collapse audit identity-like wording へ measurement error が混入する境界を先に切る必要があるため。 D02 stronger longitudinal adjunct benchmark
U0-3 threshold-stability note と overfit-exclusion rule 閾値の外的妥当性を閉じないと本人性ドリフト読解が静的一致へ戻りやすいため。 D02 session-external rerun pack と cohort extension
U14-4 filled Model Card / Dataset Card examples rerun contract を実物で埋めない限り disclosure claim が抽象論のまま残るため。 D01 adjunct rerun benchmark と leak-audit exemplars
U14-2 fixed-split rerun contract と negative-case trail exploration / confirmation split を運用文書へ落とさないと reproducibility claim が曖昧なままだからです。 D01 multi-benchmark rerun trail と split-violation examples

2026-03-30 05:07 JST addendum (current six-RQ anchor-strength recheck)

This run does not change the current public six-question batch. It only tightens how the present anchors should be read after a post-pull review of EEG-DATA, a live check of mind-upload.com/issue.html, an official-window recheck, and a read-only Todoist recheck. The operational rule stays narrow: D10 remains the only clearly public-open primary anchor for first artifacts, while D03, D02, and D01 remain subscription-gated support anchors that are still useful for the current public route but should not carry broader mechanism or field-wide reproducibility wording by themselves.

RQ Why this anchor still stays in the current batch What the current anchor is good enough to close now What must still stay outside the claim Best next adjunct before stronger wording
U13-2 D10 3M-CPSEED stays the strongest current speech-side anchor because it is public, BIDS-structured, and exposes overt, silent-intended, and imagined speech inside the same release. A brain-minus-prior control route and a narrow mimic-separation note. Preserved internal mechanism or any mind-preservation wording. An external language-evaluation layer plus a richer control family.
U13-5 D03 stays acceptable only as a bounded multimodal support anchor because it combines EEG with eye-tracking and related physiology in a subscription-gated release whose visible metadata still does not expose a recall-specific label. A perception-to-transfer branch note that stays narrow and public-safe. Any stronger mechanism-preservation reading. ChineseEEG-2 plus ZuCo 2.0 and the public robot-speech family, with SparrKULee kept as the larger auditory-only stress-test and CuBiAAD only as a later access-gated adjunct.
U0-2 D02 CSTE remains the cleanest current identity-drift floor because it is explicitly cross-session and cross-task with a biometric framing, even though the current DataPort page is subscription-gated. A state-feature collapse audit plus bounded subject-stability failure analysis. Philosophical or legal continuity claims, and stronger timing language than the dataset actually supports. A longer-horizon adjunct benchmark with more explicit timing and synchronization logs.
U0-3 The same D02 anchor keeps the threshold rule on the same cross-session / cross-task identity-drift floor as U0-2. Threshold stability and overfit-exclusion as a bounded benchmark artifact. Treating a stable threshold as a sufficient condition for personhood continuity. A session-external rerun pack and a larger cohort extension.
U14-4 D01 remains a narrow but useful disclosure anchor because a wireless-delay dataset is enough to force concrete latency and measurement fields into a Model Card / Dataset Card example. A disclosure-first rerun checklist with concrete latency fields. Any field-wide reproducibility or organizational-adoption wording. A New Benchmark Dataset Towards Ubiquitous P300 ERP-based BCI Applications as the first required public-open rerun family, then D08 for grounded validation examples.
U14-2 The same D01 route is still enough to keep the exploration/confirmation split operational and measurable at the latency-contract level. A fixed-split rerun contract plus a negative-case trail that stays narrow. Broader rerun robustness claims. The same public P300 benchmark-style family as the first required support, plus D08 for stronger split-violation grounding.

今回の再検証ログ（2026-03-29 19:07 JST）

本runでは、pull 後の current public 6RQ を再度据え置いたまま、official funding window と Todoist 実 API の「重複なく足りているか」を確認し、さらに broader funding watch が既存 task とどこまで接続済みかを整理いたしました。方針は今回も総論追加ではなく、1RQ=1検証命題=1主アンカー=1第一資金ルート=1非主張境界 を崩さないことです。

今回の再検証ログ（2026-03-29 15:15 JST）

本runでは、current public 6RQ を据え置いたまま、anchor strength と roadmap-only gap をドシエ入口へ明示いたしました。総論を増やすのではなく、「どの RQ が今の EEG-DATA で最も深く前に進みやすいか」をさらに明確にするための補足です。

今回の再検証ログ（2026-03-29 13:03 JST）

本runでは、mind-upload@12d3d42 EEG-DATA@8d3d7db889a auto-research-funds@eaa5633d0 まで pull 後に、live mind-upload.com/issue.html、主要制度の公式募集ページ、Todoist 実 API を再確認し、current public batch を据え置く判断を明文化いたしました。

今回の再検証ログ（2026-03-29 12:06 JST）

本runでは、pull 後に live mind-upload.com/issue.html を再確認し、current public batch を U13-2/U13-5/U14-4/U0-2/U0-3/U14-2 として内部ドシエの入口を同期し直しました。汎用横断の総論は増やさず、ここでも 1RQ=1検証命題=1応募テーマ=1主データ を維持いたします。

今回の再検証ログ（2026-03-29 07:18 JST）

本runでは、mind-upload.com 側でそのまま公開しても不自然になりにくい6RQを先に固定し、その6件だけを EEG-DATA 実在面・現行助成面・Todoist登録状況まで含めて再検証いたしました。汎用横断の総論は増やさず、今回も 1RQ=1検証命題=1応募テーマ=1主データ を維持いたします。

今回の再検証ログ（2026-03-29 05:55 JST）

本runでは、既存の 6RQ 深掘りキューを維持したまま、各RQを EEG-DATA の実在 bucket と auto-research-funds の現行 card 実体へ再固定いたしました。汎用横断の総論は増やさず、今回も 1RQ=1検証命題=1応募テーマ=1主データ を維持いたします。

今回の再検証ログ（2026-03-29 02:04 JST）

本runでは、現行 mind-upload の site deepening audit を強く反映する 6RQ だけを追加で固定いたしました。汎用横断の要約は増やさず、1RQ=1検証命題=1応募テーマ=1主データ を維持いたします。

今回の再検証ログ（2026-03-29 01:37 JST）

本runでは、A/B 判定の中でも「今の EEG-DATA で最初の実験と助成テーマを切りやすいもの」を優先し、可解性優先パックとして 6RQ を固定いたしました。汎用横断の要約は増やさず、1RQ=1検証命題=1応募テーマ=1主データ を継続しています。

判定内訳（再計数）

使い方（1RQずつ深掘りする運用）

  1. 対象RQの 検証命題 と 判定境界 を先に固定する。
  2. 主データ で最小再現実験を実施し、初期KPI を埋める。
  3. 停止条件 に抵触したら主張を下方修正し、応募テーマを分割する。
  4. 最低成果物 を完成させてから次RQへ進む。

今回の再検証ログ（2026-03-28 09:02 JST）

本runでは、汎用横断の更新を増やさず、1RQ=1検証命題=1応募テーマ=1主データ の方針を維持したまま、前runと重複しにくい6RQを新規の深掘り固定キューとして設定しました。

RQ-by-RQ Deep Dossiers（60件）

U0-1 同一性判定を『観測一致』と『介入応答一致』に分離したとき、どちらを必要条件・十分条件に置くか。

U0-2 時間同期（ms単位）と状態表現（行動・神経活動・生理）の対応をどう固定するか。

U0-3 同一性判定の閾値をタスク別にどう設定し、過学習モデルをどう除外するか。

U0-4 分岐/複製ケースでの『同一個体』定義をどの評価軸に帰着させるか。

U1-1 EEG/MEG逆問題での不良設定性を、どの事前分布で制御するか。

U1-2 頭蓋導電率・電極配置・ノイズ構造の不確実性を、推定不確実性へどう伝播させるか。

U1-3 同じデータで異なる逆解法（MNE, beamformer, Champagne 等）が乖離した場合の判定規則をどう置くか。

U1-4 推定値だけでなく、信頼区間/後方分布を公開基準に含めるか。

U3-1 ニューロン中心モデルに対して、グリア/体液性調節を追加した際の予測改善量をどう測るか。

U3-2 connectome完全性と機能予測性能の関係を、種横断でどう比較するか。

U3-3 身体・環境結合を除去したモデルで失われる機能をどう定量化するか。

U3-4 『必要最小構成』の判定を理論的主張ではなくデータでどう固定するか。

U3-5 glymphatic/meningeal lymphatic 系を含むとき、予測精度と説明可能性はどの程度改善するか。

U3-6 免疫監視（髄膜免疫・炎症性シグナル）を除外したモデルは、どの時点で長期予測が破綻するか。

U4-1 観測データ由来の相関を因果グラフへ持ち上げる識別条件は何か。

U4-2 介入実験（刺激・抑制・入力撹乱）で検証可能な最小因果主張は何か。

U4-3 active inferenceやDCMの理論予測を、反事実評価にどう接続するか。

U4-4 同値判定の失敗条件（falsification）をどの水準で宣言するか。

U7-1 BIDS拡張で同期・QC・刺激ログをどこまで必須化するか。

U7-2 LSL等の時刻同期誤差を検証可能な指標に落とせるか。

U7-3 アーチファクト除去（ASR, ZapLine等）の設定差が結果へ与える影響をどう監査するか。

U7-4 モダリティ間アライメント失敗時の再計測/除外基準をどう固定するか。

U7-5 前処理差分をCIで自動比較する場合、どの再現率低下をリリースブロック閾値にするか。

U7-6 モダリティ欠損（EEG欠損・fMRI欠損）条件でも同等結論を保てる最小観測セットは何か。

U8-1 閉ループBCIの遅延許容域を制御理論的にどう同定するか。

U8-2 オンライン較正と概念ドリフト対策をどう組み込むか。

U8-3 個体差と日内変動をまたぐ安定性をどの指標で評価するか。

U8-4 異常検知とフェイルセーフを評価契約へどう組み込むか。

U8-5 ヒューマンオーバーライドを導入したとき、誤作動率と回復時間をどうKPI化するか。

U8-6 週〜月スケール運用での再学習頻度を、性能劣化と安全余裕のトレードオフでどう最適化するか。

U10-1 Landauer下限を神経計算でどう適用/解釈するか。

U10-2 非平衡熱力学指標と神経情報処理効率の対応をどう定義するか。

U10-3 理論式を実データ（神経活動・代謝）へ落とし込む観測設計をどう作るか。

U10-4 WBE計算コスト評価に熱力学制約をどう統合するか。

U11-1 理論間で比較可能な入出力仕様をどう定義するか。

U11-2 PCIやIIT近似計算の計算量制約をどう扱うか。

U11-3 理論予測の対立点を単一実験計画へどう落とすか。

U11-4 意識指標を臨床/研究で運用する際の失敗条件をどう明示するか。

U12-1 分岐後主体の識別子を何に基づいて付与するか。

U12-2 責任・権利・同意の継承ルールをどの時点で分岐させるか。

U12-3 心理的連続性基準と法的個体基準の不一致をどう扱うか。

U12-4 技術評価（性能）と人格評価（帰属）をどう接続するか。

U12-5 分岐主体間で記憶編集・再同期が起きた場合、法的主体IDを再編する基準は何か。

U12-6 同意撤回が発生したとき、複数分岐主体への権限剥奪を技術的にどう実装・監査するか。

U13-1 brain-to-text成功を『意味復元』と『因果再現』へ分解できるか。

U13-2 LLMの幻覚・整合性検査を神経デコード評価へどう接続するか。

U13-3 同一出力でも内部機構が異なるケースをどう検出するか。

U13-4 模倣性能の上限を因果評価でどこまで抑制できるか。

U13-5 視覚知覚と視覚想起で同一デコーダを使ったとき、意味復元精度の劣化パターンはどこで分岐するか。

U13-6 プロンプト誘導・データリーク・shortcut学習を分離検出する対照実験をどう設計するか。

U14-1 データ/コード/評価環境の固定粒度をどこまで要求するか。

U14-2 探索研究と検証研究を運用上どう分離するか。

U14-3 leaderboardでのリーク・過適合・報告バイアスをどう監査するか。

U14-4 Model Card / Dataset Card を評価契約へどう統合するか。

U14-5 再現失敗ケースを否定例レジストリとして公開し、再試行サイクルをどう運用するか。

U14-6 コンテナ固定（OS・依存ライブラリ・乱数種）を必須化した場合、再現コスト増分をどこまで許容するか。

U15-1 神経データの機微性をどの法概念で扱うか（個人情報・生体情報・人格情報）。

U15-2 neurorightsを技術監査項目へどう写像するか。

U15-3 法域差（EU/US/JP等）を跨ぐ最小共通運用をどう定義するか。

U15-4 技術進展に応じた停止条件・更新条件をどうガバナンス化するか。

運用メモ

2026-03-26 12:03 JST 再検証ログ（本run / 1RQずつ深掘りキュー再固定）

2026-03-27 09:30 JST 再検証ログ（本run / 非重複6RQの深掘り固定）

2026-03-28 07:02 JST 再検証ログ（本run / 非重複6RQの深掘り固定）

2026-03-28 08:02 JST 再検証ログ（本run / 非重複6RQの深掘り固定）

今回の再検証ログ（2026-03-28 16:02 JST）

本runでは、汎用横断サマリを増やさず、1RQ=1検証命題=1応募テーマ=1主データ の実行粒度を維持したまま、深掘り対象を次の6RQに固定しました。