Sync GitHub main e8a04aea24121d15707238534e635cf0ceee8eaf (part 3)

research/generated/CATALOG_AUDIT_REPORT.json

@@ -2,8 +2,8 @@
   "status": "clean",
   "audit": {
     "clean": false,
-    "question_reference_count": 5000,
-    "hypothesis_reference_count": 3661,
+    "question_reference_count": 5024,
+    "hypothesis_reference_count": 3709,
     "missing_questions": [
       "RQ-AI-DESIGN-001",
       "RQ-AI-EFF-001",

research/generated/CATALOG_AUDIT_REPORT.md

@@ -5,8 +5,8 @@
 The canonical ResearchRegistry remains authoritative. Historical/design references and test-only fixtures are listed with explicit reasons; an unknown missing identifier fails CI.
 
 ## Summary
-- Question references: 5000
-- Hypothesis references: 3661
+- Question references: 5024
+- Hypothesis references: 3709
 - Missing questions: 27
 - Missing hypotheses: 24
 - Registry link issues: 0

research/generated/CLAIM_REGISTER.md

@@ -1,6 +1,6 @@
 # MHRN Claim Register
 
-*Generiert am 2026-09-21*
+*Generiert am 2026-09-27*
 
 | Claim | Status | Konfidenz | Evidenzen | Experimente |
 |-------|--------|-----------|-----------|-------------|
@@ -13,6 +13,9 @@
 | `CLAIM-SELF-001`: In Brain-5D entstehen spontan funktionale Module ohne explizite Programmierung.... | ⬜ untested | none | 0 | 0 |
 | `CLAIM-AIR-001`: Ein standardisiertes ResearchPacket verbessert die Identifikation vorab definier... | ⬜ untested | none | 0 | 0 |
 | `CLAIM-EVAL-006`: Unter dem eingefrorenen V2-Protokoll konformieren die deklarierten MHRN-Einzelze... | 🔄 inconclusive | low | 1 | 2 |
+| `CLAIM-S1-TOPO-001`: Unter kontrollierter Neuronenzahl, globalem Kantenbudget, Stimulus, Gewichten un... | 🔄 inconclusive | low | 1 | 3 |
+| `CLAIM-S1-EFFICIENCY-001`: Im registrierten 64-Neuronen-Small-SNN-Regime kann eine spezifische 5D-Koordinat... | ⬜ untested | none | 0 | 0 |
+| `CLAIM-S1-TEMP-001`: Unter dem registrierten sechsneuronigen Small-SNN-Task erzeugen unterscheidbare ... | 🔄 inconclusive | low | 1 | 2 |
 
 
 > Pruefstatus: RQ/H- und EVID-Statuswerte geben den Registry-Inhalt wieder. Insbesondere historische supports/supported-Eintraege sind keine Bestaetigung einer Freigabe nach den heutigen Clean-Freeze- und Human-Review-Gates. Ein abgeschlossener Lauf ist DATA, nicht automatisch akzeptierte Evidenz.

research/generated/CROSSWALK.md

@@ -96,7 +96,7 @@ All registered RQs are projected. Empty direction coordinates remain explicit un
 | RQ-SELF-002 | unresolved | unclassified | unresolved | unresolved | H-SELF-002-A | EXP-BATCH-20260906200118-25, EXP-BATCH-20260908200906-25, EXP-BATCH-20260909223705-25, EXP-BATCH-20260914074039-25 | unresolved | unresolved |
 | RQ-SNN-001 | unresolved | unclassified | unresolved | unresolved | H-SNN-001-A | EXP-2026-0001, EXP-BATCH-20260906195229-01, EXP-BATCH-20260906195506-01, EXP-BATCH-20260906195731-01, EXP-BATCH-20260906200118-01, EXP-BATCH-20260908200906-01, EXP-BATCH-20260914074039-01, EXP-EMP-20260910, EXP-GEN-0001, EXP-GEN-0002, EXP-GEN-0008, EXP-GEN-0010, EXP-GEN-0017, EXP-GEN-0019, EXP-GEN-0020, EXP-GEN-0021, EXP-GEN-0027, EXP-GEN-0028, EXP-GEN-0030, EXP-GEN-0031, EXP-GEN-0032, EXP-GEN-0037, EXP-GEN-0038, EXP-GEN-0039, EXP-GEN-0040, EXP-GEN-0042, EXP-GEN-0043, EXP-GEN-0044, EXP-SNN-001, EXP-SNN-001-R1, EXP-SNN-001-R2, EXP-SNN-001-R3, EXP-SNN-001-R5, EXP-STDP-0001a | unresolved | unresolved |
 | RQ-SNN-002 | unresolved | unclassified | unresolved | III, IV | H-SNN-002-A | EXP-BATCH-20260906200118-02, EXP-BATCH-20260908200906-02, EXP-BATCH-20260914074039-02, EXP-GEN-0003, EXP-GEN-0011, EXP-GEN-0012, EXP-GEN-0022, EXP-GEN-0026, EXP-GEN-0029, EXP-GEN-0034, EXP-GEN-0035, EXP-GEN-0045, EXP-SNN-002-R1, EXP-SNN-002-R2 | unresolved | unresolved |
-| RQ-SNN-003 | unresolved | unclassified | unresolved | III, IV, X, XI | H-SNN-003-B | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | unresolved | research/publications/papers/2026-09-20_topology-dynamics_v0.1/PAPER.md |
+| RQ-SNN-003 | unresolved | unclassified | unresolved | IV, XI | H-SNN-003-B, H-SNN-003-C | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-PROMO-R1-20260927, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | EVID-2026-19 | research/publications/papers/2026-09-20_topology-dynamics_v0.1/PAPER.md |
 | RQ-SNN-004 | unresolved | unclassified | unresolved | IV | H-SNN-004-A | EXP-BATCH-20260906200118-05, EXP-BATCH-20260908200906-05, EXP-BATCH-20260909223705-05, EXP-BATCH-20260914074039-05, EXP-SNN004-STDP-ASYM-R2-20260919 | unresolved | unresolved |
 | RQ-SNN-005 | unresolved | unclassified | unresolved | IV | H-SNN-005-A | EXP-BATCH-20260906200118-06, EXP-BATCH-20260908200906-06, EXP-BATCH-20260909223705-06, EXP-BATCH-20260914074039-06, EXP-LEARN-20260903-2300, EXP-LEARN-20260903-2330, EXP-SNN005-HOLDOUT-R2-20260919 | unresolved | unresolved |
 | RQ-SNN-006 | unresolved | unclassified | unresolved | unresolved | H-SNN-006-A | EXP-GEN-0041 | unresolved | unresolved |
@@ -109,7 +109,7 @@ All registered RQs are projected. Empty direction coordinates remain explicit un
 | RQ-STRUCT-001 | unresolved | unclassified | unresolved | unresolved | H-STRUCT-001-A | EXP-BATCH-20260906200118-26, EXP-BATCH-20260908200906-26, EXP-BATCH-20260909223705-26, EXP-BATCH-20260914074039-26 | unresolved | unresolved |
 | RQ-SUITE-001 | unresolved | unclassified | unresolved | unresolved | H-SUITE-001-A | EXP-BATCH-20260906200118-34, EXP-BATCH-20260909223705-34, EXP-BATCH-20260914074039-34, EXP-GEN-0024, EXP-GEN-0025, EXP-GEN-0033, EXP-GEN-0033-R1, EXP-GEN-0036 | unresolved | unresolved |
 | RQ-TEMP-001 | unresolved | unclassified | unresolved | unresolved | H-TEMP-001-A | EXP-BATCH-20260906200118-08, EXP-BATCH-20260908200906-08, EXP-BATCH-20260909223705-08, EXP-BATCH-20260914074039-08, EXP-GEN-0013, EXP-GEN-0014, EXP-GEN-0015, EXP-GEN-0016, EXP-GEN-0018, EXP-TEMP-0001 | unresolved | unresolved |
-| RQ-TEMP-002 | unresolved | unclassified | unresolved | III, IV | H-TEMP-002-A | EXP-BATCH-20260906200118-40, EXP-BATCH-20260909223705-40, EXP-BATCH-20260914074039-40, EXP-EMP-20260910, EXP-S1-TEMP-ORDER-V2-20260919 | unresolved | unresolved |
+| RQ-TEMP-002 | unresolved | unclassified | unresolved | III, IV | H-TEMP-002-A | EXP-BATCH-20260906200118-40, EXP-BATCH-20260909223705-40, EXP-BATCH-20260914074039-40, EXP-EMP-20260910, EXP-S1-TEMP-ORDER-V2-20260919, EXP-S1-TEMP-PROMO-R1-20260927 | EVID-2026-20 | unresolved |
 | RQ-TIME-001 | unresolved | unclassified | unresolved | unresolved | H-TIME-001-A | EXP-BATCH-20260906200118-09, EXP-BATCH-20260908200906-09, EXP-BATCH-20260909223705-09, EXP-BATCH-20260914074039-09, EXP-TIME-0001 | unresolved | unresolved |
 | RQ-TIME-002 | unresolved | unclassified | unresolved | unresolved | H-TIME-002-A | EXP-BATCH-20260909223705-77, EXP-BATCH-20260914074039-77, EXP-EMP-20260910 | unresolved | unresolved |
 | RQ-WEL-101 | unresolved | unclassified | unresolved | VIII | H-WEL-101-A | unresolved | unresolved | unresolved |

research/generated/DISSERTATION_MAP.md

@@ -28,8 +28,7 @@ Source: `research/publications/2026-09-17_recursive-epistemics_v1.8/parts/03_res
 | RQ-DET-001 | open | unresolved | unclassified | EXP-2026-0001, EXP-BATCH-20260906200118-03, EXP-BATCH-20260908200906-03, EXP-BATCH-20260914074039-03, EXP-DET-0001, EXP-GEN-0023, EXP-GEN-0046 | EVID-2026-01, EVID-2026-03, EVID-2026-05, EVID-2026-07, EVID-2026-09, EVID-2026-11, EVID-2026-13, EVID-2026-15 |
 | RQ-REC-002 | open | unresolved | unclassified | EXP-BATCH-20260906200118-42, EXP-BATCH-20260909223705-42, EXP-BATCH-20260914074039-42, EXP-EMP-20260910, EXP-REC-002-CLEAN-R2-20260919 | unresolved |
 | RQ-SNN-002 | open | unresolved | unclassified | EXP-BATCH-20260906200118-02, EXP-BATCH-20260908200906-02, EXP-BATCH-20260914074039-02, EXP-GEN-0003, EXP-GEN-0011, EXP-GEN-0012, EXP-GEN-0022, EXP-GEN-0026, EXP-GEN-0029, EXP-GEN-0034, EXP-GEN-0035, EXP-GEN-0045, EXP-SNN-002-R1, EXP-SNN-002-R2 | unresolved |
-| RQ-SNN-003 | open | unresolved | unclassified | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | unresolved |
-| RQ-TEMP-002 | open | unresolved | unclassified | EXP-BATCH-20260906200118-40, EXP-BATCH-20260909223705-40, EXP-BATCH-20260914074039-40, EXP-EMP-20260910, EXP-S1-TEMP-ORDER-V2-20260919 | unresolved |
+| RQ-TEMP-002 | inconclusive | unresolved | unclassified | EXP-BATCH-20260906200118-40, EXP-BATCH-20260909223705-40, EXP-BATCH-20260914074039-40, EXP-EMP-20260910, EXP-S1-TEMP-ORDER-V2-20260919, EXP-S1-TEMP-PROMO-R1-20260927 | EVID-2026-20 |
 
 ## Teil IV - Empirisches Forschungsprogramm
 
@@ -45,10 +44,10 @@ Source: `research/publications/2026-09-17_recursive-epistemics_v1.8/parts/04_emp
 | RQ-GW-006 | open | unresolved | unclassified | EXP-BATCH-20260909223705-83, EXP-BATCH-20260914074039-88 | unresolved |
 | RQ-MEM-001 | open | unresolved | unclassified | EXP-BATCH-20260906200118-27, EXP-BATCH-20260909223705-27, EXP-BATCH-20260914074039-27 | unresolved |
 | RQ-SNN-002 | open | unresolved | unclassified | EXP-BATCH-20260906200118-02, EXP-BATCH-20260908200906-02, EXP-BATCH-20260914074039-02, EXP-GEN-0003, EXP-GEN-0011, EXP-GEN-0012, EXP-GEN-0022, EXP-GEN-0026, EXP-GEN-0029, EXP-GEN-0034, EXP-GEN-0035, EXP-GEN-0045, EXP-SNN-002-R1, EXP-SNN-002-R2 | unresolved |
-| RQ-SNN-003 | open | unresolved | unclassified | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | unresolved |
+| RQ-SNN-003 | inconclusive | unresolved | unclassified | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-PROMO-R1-20260927, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | EVID-2026-19 |
 | RQ-SNN-004 | open | unresolved | unclassified | EXP-BATCH-20260906200118-05, EXP-BATCH-20260908200906-05, EXP-BATCH-20260909223705-05, EXP-BATCH-20260914074039-05, EXP-SNN004-STDP-ASYM-R2-20260919 | unresolved |
 | RQ-SNN-005 | open | unresolved | unclassified | EXP-BATCH-20260906200118-06, EXP-BATCH-20260908200906-06, EXP-BATCH-20260909223705-06, EXP-BATCH-20260914074039-06, EXP-LEARN-20260903-2300, EXP-LEARN-20260903-2330, EXP-SNN005-HOLDOUT-R2-20260919 | unresolved |
-| RQ-TEMP-002 | open | unresolved | unclassified | EXP-BATCH-20260906200118-40, EXP-BATCH-20260909223705-40, EXP-BATCH-20260914074039-40, EXP-EMP-20260910, EXP-S1-TEMP-ORDER-V2-20260919 | unresolved |
+| RQ-TEMP-002 | inconclusive | unresolved | unclassified | EXP-BATCH-20260906200118-40, EXP-BATCH-20260909223705-40, EXP-BATCH-20260914074039-40, EXP-EMP-20260910, EXP-S1-TEMP-ORDER-V2-20260919, EXP-S1-TEMP-PROMO-R1-20260927 | EVID-2026-20 |
 
 ## Teil V - Wissenschaftliche Infrastruktur und Engineering
 
@@ -129,7 +128,6 @@ Source: `research/publications/2026-09-17_recursive-epistemics_v1.8/parts/10_syn
 | --- | --- | --- | --- | --- | --- |
 | RQ-5D-005 | open | unresolved | unclassified | EXP-BATCH-20260906200118-38, EXP-BATCH-20260909223705-38, EXP-BATCH-20260914074039-38, EXP-EMP-20260910 | unresolved |
 | RQ-DET-001 | open | unresolved | unclassified | EXP-2026-0001, EXP-BATCH-20260906200118-03, EXP-BATCH-20260908200906-03, EXP-BATCH-20260914074039-03, EXP-DET-0001, EXP-GEN-0023, EXP-GEN-0046 | EVID-2026-01, EVID-2026-03, EVID-2026-05, EVID-2026-07, EVID-2026-09, EVID-2026-11, EVID-2026-13, EVID-2026-15 |
-| RQ-SNN-003 | open | unresolved | unclassified | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | unresolved |
 
 ## Teil XI - Offene Forschungslandschaft
 
@@ -144,7 +142,7 @@ Source: `research/publications/2026-09-17_recursive-epistemics_v1.8/parts/11_ope
 | RQ-GW-006 | open | unresolved | unclassified | EXP-BATCH-20260909223705-83, EXP-BATCH-20260914074039-88 | unresolved |
 | RQ-META-003 | open | unresolved | epistemological_methodological | unresolved | unresolved |
 | RQ-META-005 | open | unresolved | epistemological_methodological | unresolved | unresolved |
-| RQ-SNN-003 | open | unresolved | unclassified | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | unresolved |
+| RQ-SNN-003 | inconclusive | unresolved | unclassified | EXP-BATCH-20260906200118-04, EXP-BATCH-20260908200906-04, EXP-BATCH-20260909223705-04, EXP-BATCH-20260914074039-04, EXP-GEN-0047, EXP-S1-TOPO-PROMO-R1-20260927, EXP-S1-TOPO-V2-20260918, EXP-S1-TOPO-V3-20260918, EXP-S1-TOPO-V3-R1-20260918 | EVID-2026-19 |
 
 ## Unmapped registered RQs
 

research/generated/EVIDENCE_MATRIX.md

@@ -1,6 +1,6 @@
 # MHRN Evidence Matrix
 
-*Generiert am 2026-09-21*
+*Generiert am 2026-09-27*
 
 RQ-Status und Claim-Status sind unterschiedliche wissenschaftliche Zustände und werden nicht gegenseitig abgeleitet.
 
@@ -9,7 +9,7 @@ RQ-Status und Claim-Status sind unterschiedliche wissenschaftliche Zustände und
 | `RQ-SNN-001` | open | `H-SNN-001-A` | — | — | 2 | `EXP-2026-0001`, `EXP-BATCH-20260906195229-01`, `EXP-BATCH-20260906195506-01`, `EXP-BATCH-20260906195731-01`, `EXP-BATCH-20260906200118-01`, `EXP-BATCH-20260908200906-01`, `EXP-BATCH-20260914074039-01`, `EXP-EMP-20260910`, `EXP-GEN-0001`, `EXP-GEN-0002`, `EXP-GEN-0008`, `EXP-GEN-0010`, `EXP-GEN-0017`, `EXP-GEN-0019`, `EXP-GEN-0020`, `EXP-GEN-0021`, `EXP-GEN-0027`, `EXP-GEN-0028`, `EXP-GEN-0030`, `EXP-GEN-0031`, `EXP-GEN-0032`, `EXP-GEN-0037`, `EXP-GEN-0038`, `EXP-GEN-0039`, `EXP-GEN-0040`, `EXP-GEN-0042`, `EXP-GEN-0043`, `EXP-GEN-0044`, `EXP-SNN-001`, `EXP-SNN-001-R1`, `EXP-SNN-001-R2`, `EXP-SNN-001-R3`, `EXP-SNN-001-R5`, `EXP-STDP-0001a` | — | offen |
 | `RQ-SNN-002` | open | `H-SNN-002-A` | — | — | 2 | `EXP-BATCH-20260906200118-02`, `EXP-BATCH-20260908200906-02`, `EXP-BATCH-20260914074039-02`, `EXP-GEN-0003`, `EXP-GEN-0011`, `EXP-GEN-0012`, `EXP-GEN-0022`, `EXP-GEN-0026`, `EXP-GEN-0029`, `EXP-GEN-0034`, `EXP-GEN-0035`, `EXP-GEN-0045`, `EXP-SNN-002-R1`, `EXP-SNN-002-R2` | — | offen |
 | `RQ-DET-001` | open | `H-SNN-003-A` | `CLAIM-DET-001` | `CLAIM-DET-001`=inconclusive | 1 | `EXP-2026-0001`, `EXP-BATCH-20260906200118-03`, `EXP-BATCH-20260908200906-03`, `EXP-BATCH-20260914074039-03`, `EXP-DET-0001`, `EXP-GEN-0023`, `EXP-GEN-0046` | `EVID-2026-01`, `EVID-2026-03`, `EVID-2026-05`, `EVID-2026-07`, `EVID-2026-09`, `EVID-2026-11`, `EVID-2026-13`, `EVID-2026-15` | offen |
-| `RQ-SNN-003` | open | `H-SNN-003-B` | — | — | 2 | `EXP-BATCH-20260906200118-04`, `EXP-BATCH-20260908200906-04`, `EXP-BATCH-20260909223705-04`, `EXP-BATCH-20260914074039-04`, `EXP-GEN-0047`, `EXP-S1-TOPO-V2-20260918`, `EXP-S1-TOPO-V3-20260918`, `EXP-S1-TOPO-V3-R1-20260918` | — | replicated_internal_data_human_reviewed_evid_not_promoted |
+| `RQ-SNN-003` | inconclusive | `H-SNN-003-B`, `H-SNN-003-C` | `CLAIM-S1-TOPO-001`, `CLAIM-S1-EFFICIENCY-001` | `CLAIM-S1-TOPO-001`=inconclusive, `CLAIM-S1-EFFICIENCY-001`=untested | 3 | `EXP-BATCH-20260906200118-04`, `EXP-BATCH-20260908200906-04`, `EXP-BATCH-20260909223705-04`, `EXP-BATCH-20260914074039-04`, `EXP-GEN-0047`, `EXP-S1-TOPO-PROMO-R1-20260927`, `EXP-S1-TOPO-V2-20260918`, `EXP-S1-TOPO-V3-20260918`, `EXP-S1-TOPO-V3-R1-20260918` | `EVID-2026-19` | bounded_topology_evid_efficiency_hypothesis_untested |
 | `RQ-SNN-004` | open | `H-SNN-004-A` | `CLAIM-SNN-001` | `CLAIM-SNN-001`=untested | 2 | `EXP-BATCH-20260906200118-05`, `EXP-BATCH-20260908200906-05`, `EXP-BATCH-20260909223705-05`, `EXP-BATCH-20260914074039-05`, `EXP-SNN004-STDP-ASYM-R2-20260919` | — | offen |
 | `RQ-SNN-005` | open | `H-SNN-005-A` | — | — | 2 | `EXP-BATCH-20260906200118-06`, `EXP-BATCH-20260908200906-06`, `EXP-BATCH-20260909223705-06`, `EXP-BATCH-20260914074039-06`, `EXP-LEARN-20260903-2300`, `EXP-LEARN-20260903-2330`, `EXP-SNN005-HOLDOUT-R2-20260919` | — | offen |
 | `RQ-PING-001` | open | `H-PING-001-A` | — | — | 0 | `EXP-BATCH-20260906200118-07`, `EXP-BATCH-20260908200906-07`, `EXP-BATCH-20260909223705-07`, `EXP-BATCH-20260914074039-07`, `EXP-GEN-0006`, `EXP-GEN-0007`, `EXP-GEN-0009`, `EXP-PING-0001`, `EXP-PING-0001-v2`, `EXP-PING-0001-v3` | — | offen |
@@ -46,7 +46,7 @@ RQ-Status und Claim-Status sind unterschiedliche wissenschaftliche Zustände und
 | `RQ-REPL-001` | open | `H-REPL-001-A` | — | — | 0 | `EXP-BATCH-20260909223705-37`, `EXP-BATCH-20260914074039-37`, `EXP-EMP-20260910`, `EXP-REPL-0001-R1`, `EXP-RETRY-AIRR-2026-0001-01` | — | offen |
 | `RQ-5D-005` | open | `H-5D-005-A` | — | — | 0 | `EXP-BATCH-20260906200118-38`, `EXP-BATCH-20260909223705-38`, `EXP-BATCH-20260914074039-38`, `EXP-EMP-20260910` | — | offen |
 | `RQ-REG-002` | open | `H-REG-002-A`, `H-REG-002-B` | — | — | 0 | `EXP-BATCH-20260906200118-39`, `EXP-BATCH-20260909223705-39`, `EXP-BATCH-20260914074039-39`, `EXP-EMP-20260910`, `EXP-REG-0002-R1` | — | offen |
-| `RQ-TEMP-002` | open | `H-TEMP-002-A` | — | — | 0 | `EXP-BATCH-20260906200118-40`, `EXP-BATCH-20260909223705-40`, `EXP-BATCH-20260914074039-40`, `EXP-EMP-20260910`, `EXP-S1-TEMP-ORDER-V2-20260919` | — | preregistered_task_data_human_reviewed_evid_not_promoted |
+| `RQ-TEMP-002` | inconclusive | `H-TEMP-002-A` | `CLAIM-S1-TEMP-001` | `CLAIM-S1-TEMP-001`=inconclusive | 0 | `EXP-BATCH-20260906200118-40`, `EXP-BATCH-20260909223705-40`, `EXP-BATCH-20260914074039-40`, `EXP-EMP-20260910`, `EXP-S1-TEMP-ORDER-V2-20260919`, `EXP-S1-TEMP-PROMO-R1-20260927` | `EVID-2026-20` | bounded_temporal_order_evid_with_methodological_limitations |
 | `RQ-PERF-001` | open | `H-PERF-001-A` | — | — | 0 | `EXP-BATCH-20260906200118-41`, `EXP-BATCH-20260909223705-41`, `EXP-BATCH-20260914074039-41`, `EXP-EMP-20260910` | — | offen |
 | `RQ-REC-002` | open | `H-REC-002-A` | — | — | 0 | `EXP-BATCH-20260906200118-42`, `EXP-BATCH-20260909223705-42`, `EXP-BATCH-20260914074039-42`, `EXP-EMP-20260910`, `EXP-REC-002-CLEAN-R2-20260919` | — | offen |
 | `RQ-LIFE-001` | open | `H-LIFE-001-A` | — | — | 0 | `EXP-BATCH-20260909223705-43`, `EXP-BATCH-20260914074039-43`, `EXP-EMP-20260910`, `EXP-LIFE-0001-R1`, `EXP-RETRY-AIRR-2026-0001-02` | — | offen |
@@ -127,17 +127,17 @@ RQ-Status und Claim-Status sind unterschiedliche wissenschaftliche Zustände und
 |-----------|--------|
 | answered | 1 |
 | in_progress | 1 |
-| inconclusive | 1 |
-| open | 109 |
+| inconclusive | 3 |
+| open | 107 |
 | **Gesamt RQs** | **112** |
 
 ### Claims (Claim-Status)
 
 | Claim-Status | Anzahl |
 |--------------|--------|
-| inconclusive | 3 |
-| untested | 6 |
-| **Gesamt Claims** | **9** |
+| inconclusive | 5 |
+| untested | 7 |
+| **Gesamt Claims** | **12** |
 
 ---
 *Automatisch generiert — RQ-Status, Claim-Status, DATA und EVID bleiben getrennte Ebenen.*

research/generated/LITERATURE_MATRIX.md

@@ -10,8 +10,10 @@
 | Markram (2015) | Large-scale SNN simulations can reproduce microcircuit dynamics. | RQ-SCALE-001 | open |
 | Maass (1997) | Spiking neurons are computationally more powerful than sigmoidal neurons. | RQ-SNN-001 | open |
 | Hebb (1949) | Neurons that fire together, wire together. | RQ-SNN-004, RQ-SELF-001 | open |
-| Watts (1998) | Small-world networks combine high clustering with short path lengths. | RQ-SNN-003 | open |
-| Barabási (1999) | Preferential attachment produces scale-free network degree distributions. | RQ-SNN-003 | open |
+| Watts (1998) | Small-world networks combine high clustering with short path lengths. | RQ-SNN-003 | inconclusive |
+| Barabási (1999) | Preferential attachment produces scale-free network degree distributions. | RQ-SNN-003 | inconclusive |
+| Olshausen (1996) | Sparse coding is a biologically motivated efficient-coding principle; this source does not imply that MHRN Stage-1 topology is a sparse code. | RQ-SNN-003 | inconclusive |
+| Stimberg (2019) | Brian 2 is a flexible simulator for spiking neural networks and is used in MHRN only as an external reference implementation where explicitly declared. | RQ-EVAL-006, RQ-SNN-003 | inconclusive |
 | others (2026) |  | RQ-CNS-101, RQ-CNS-112, RQ-CNS-113, RQ-EPI-102, RQ-WEL-102 | open |
 | others (2025) |  | RQ-CNS-101, RQ-EPI-101 | open |
 | Naccache (2025) |  | RQ-CNS-101 | open |
@@ -64,4 +66,4 @@
 | Langosco (2022) |  | RQ-SAFE-002 | open |
 
 ---
-*Automatisch generiert am 2026-09-21*
+*Automatisch generiert am 2026-09-27*

research/generated/META_RESEARCH_CROSSWALK.json

@@ -2803,37 +2803,47 @@
       "stages": []
     },
     {
-      "claim_statuses": {},
-      "claims": [],
+      "claim_statuses": {
+        "CLAIM-S1-EFFICIENCY-001": "untested",
+        "CLAIM-S1-TOPO-001": "inconclusive"
+      },
+      "claims": [
+        "CLAIM-S1-EFFICIENCY-001",
+        "CLAIM-S1-TOPO-001"
+      ],
       "cross_cuts": [],
       "direction": null,
       "direction_status": "unresolved_no_inference",
       "domain": "Spiking Neural Networks",
-      "evidence_records": [],
+      "evidence_records": [
+        "EVID-2026-19"
+      ],
       "experiments": [
         "EXP-BATCH-20260906200118-04",
         "EXP-BATCH-20260908200906-04",
         "EXP-BATCH-20260909223705-04",
         "EXP-BATCH-20260914074039-04",
         "EXP-GEN-0047",
+        "EXP-S1-TOPO-PROMO-R1-20260927",
         "EXP-S1-TOPO-V2-20260918",
         "EXP-S1-TOPO-V3-20260918",
         "EXP-S1-TOPO-V3-R1-20260918"
       ],
       "hypotheses": [
-        "H-SNN-003-B"
+        "H-SNN-003-B",
+        "H-SNN-003-C"
       ],
       "hypothesis_statuses": {
-        "H-SNN-003-B": "untested"
+        "H-SNN-003-B": "inconclusive",
+        "H-SNN-003-C": "untested"
       },
       "literature": [
         "SRC-WATTS-STROGATZ-1998",
-        "SRC-BARABASI-1999"
+        "SRC-BARABASI-1999",
+        "SRC-OLSHAUSEN-FIELD-1996"
       ],
       "manuscript_parts": [
-        "III",
         "IV",
-        "X",
         "XI"
       ],
       "papers": [
@@ -2841,7 +2851,7 @@
       ],
       "research_axis": "unclassified",
       "research_question": "RQ-SNN-003",
-      "rq_status": "open",
+      "rq_status": "inconclusive",
       "source_path": "research/registry/questions.yaml",
       "stages": []
     },
@@ -3248,25 +3258,32 @@
       "stages": []
     },
     {
-      "claim_statuses": {},
-      "claims": [],
+      "claim_statuses": {
+        "CLAIM-S1-TEMP-001": "inconclusive"
+      },
+      "claims": [
+        "CLAIM-S1-TEMP-001"
+      ],
       "cross_cuts": [],
       "direction": null,
       "direction_status": "unresolved_no_inference",
       "domain": "Temporal Learning",
-      "evidence_records": [],
+      "evidence_records": [
+        "EVID-2026-20"
+      ],
       "experiments": [
         "EXP-BATCH-20260906200118-40",
         "EXP-BATCH-20260909223705-40",
         "EXP-BATCH-20260914074039-40",
         "EXP-EMP-20260910",
-        "EXP-S1-TEMP-ORDER-V2-20260919"
+        "EXP-S1-TEMP-ORDER-V2-20260919",
+        "EXP-S1-TEMP-PROMO-R1-20260927"
       ],
       "hypotheses": [
         "H-TEMP-002-A"
       ],
       "hypothesis_statuses": {
-        "H-TEMP-002-A": "untested"
+        "H-TEMP-002-A": "inconclusive"
       },
       "literature": [],
       "manuscript_parts": [
@@ -3276,7 +3293,7 @@
       "papers": [],
       "research_axis": "unclassified",
       "research_question": "RQ-TEMP-002",
-      "rq_status": "open",
+      "rq_status": "inconclusive",
       "source_path": "research/registry/questions.yaml",
       "stages": []
     },
@@ -5397,6 +5414,28 @@
         "research/experiments/EXP-S1-TEMP-ORDER-V2-20260919/review_request.json.review.json"
       ]
     },
+    {
+      "authority": "review_record_only_not_EVID_or_replication",
+      "experiment_id": "EXP-S1-TEMP-PROMO-R1-20260927",
+      "manifest_review_status": "unknown",
+      "reconciliation_state": "recorded_review_present_check_status_binding",
+      "review_artifacts": [
+        "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/human_review.json",
+        "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/review_request.json",
+        "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/review_request.json.review.json"
+      ]
+    },
+    {
+      "authority": "review_record_only_not_EVID_or_replication",
+      "experiment_id": "EXP-S1-TOPO-PROMO-R1-20260927",
+      "manifest_review_status": "unknown",
+      "reconciliation_state": "recorded_review_present_check_status_binding",
+      "review_artifacts": [
+        "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/human_review.json",
+        "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/review_request.json",
+        "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/review_request.json.review.json"
+      ]
+    },
     {
       "authority": "review_record_only_not_EVID_or_replication",
       "experiment_id": "EXP-S1-TOPO-V2-20260918",

research/generated/OPEN_QUESTIONS.md

@@ -1,6 +1,6 @@
 # MHRN Open Questions
 
-*Generiert am 2026-09-21*
+*Generiert am 2026-09-27*
 
 Die folgenden Forschungsfragen sind noch offen und warten auf experimentelle Evidenz.
 
@@ -65,9 +65,11 @@ Die folgenden Forschungsfragen sind noch offen und warten auf experimentelle Evi
 **Literatur:**
 - `SRC-WATTS-STROGATZ-1998`: Duncan J. Watts et al. (1998)
 - `SRC-BARABASI-1999`: Albert-László Barabási et al. (1999)
+- `SRC-OLSHAUSEN-FIELD-1996`: Bruno A. Olshausen et al. (1996)
 
 **Hypothesen:**
 - `H-SNN-003-B`: Unter gleicher Neuronenzahl, Dichte und Stimulusbedingung unterscheiden sich Propagationslatenz oder -reichweite zwischen mindestens zwei vorab definierten Topologien; ein Vorteil einer 5D-Anordnung wird nicht vorausgesetzt.
+- `H-SNN-003-C`: Unter dem registrierten 64-Neuronen-Small-SNN-Regime zeigt die 5D-Koordinatenorganisation nach Kontrolle der dynamischen Rekrutierung eine höhere Propagations-Effizienz als vorab definierte 3D-, 5D-shuffled- und Random-Graph-Kontrollen, operationalisiert als activation_auc_0_32 pro Spike und pro geliefertem Event; ein allgemeiner 5D-Vorteil wird nicht behauptet.
 
 ---
 

research/generated/RESEARCH_CATALOG.md

@@ -1,13 +1,13 @@
 # MHRN Research Catalog
 
-*Generiert am 2026-09-21*
+*Generiert am 2026-09-27*
 
 ## Übersicht
 
 - **Forschungsfragen:** 112
-- **Hypothesen:** 120
-- **Claims:** 9
-- **Literaturquellen:** 60
+- **Hypothesen:** 121
+- **Claims:** 12
+- **Literaturquellen:** 62
 
 ---
 
@@ -1549,21 +1549,29 @@
 
 **Frage:** Wie variiert die Propagation mit der Topologie?
 
-**Status:** open
+**Status:** inconclusive
 **Relevanz:** Grundlegendes Verständnis der Signalausbreitung in multidimensionalen SNNs.
 
 **Aktuelle Antwort:**
 
-> EXP-S1-TOPO-V2-20260918 und EXP-S1-TOPO-V3-R1-20260918 bilden gemeinsam die kanonische Stage-1-Topologie-DATA-Linie. Beide Human Reviews durch Thomas Heisig akzeptieren die begrenzte Interpretation. Die DATA sind nicht als EVID promotet.
+> CLAIM-S1-TOPO-001 besitzt seit EXP-S1-TOPO-PROMO-R1-20260927 kanonische EVID (EVID-2026-19) fuer die eng begrenzte Aussage, dass Topologie im registrierten 64-Neuronen-/246-Kanten-Regime Propagationsmetriken veraendert. Eine nachtraeglich beobachtete AUC-pro-Spike/-Event-Systematik wird getrennt als explorative Motivation fuer H-SNN-003-C gefuehrt und ist keine EVID.
 
-*Konfidenz: replicated_internal_data_human_reviewed_evid_not_promoted*
+*Konfidenz: bounded_topology_evid_efficiency_hypothesis_untested*
 
 **Hypothesen:**
-- `H-SNN-003-B`: Unter gleicher Neuronenzahl, Dichte und Stimulusbedingung unterscheiden sich Propagationslatenz oder -reichweite zwischen mindestens zwei vorab definierten Topologien; ein Vorteil einer 5D-Anordnung wird nicht vorausgesetzt. *(untested)*
+- `H-SNN-003-B`: Unter gleicher Neuronenzahl, Dichte und Stimulusbedingung unterscheiden sich Propagationslatenz oder -reichweite zwischen mindestens zwei vorab definierten Topologien; ein Vorteil einer 5D-Anordnung wird nicht vorausgesetzt. *(inconclusive)* — Evidenz: EVID-2026-19
+- `H-SNN-003-C`: Unter dem registrierten 64-Neuronen-Small-SNN-Regime zeigt die 5D-Koordinatenorganisation nach Kontrolle der dynamischen Rekrutierung eine höhere Propagations-Effizienz als vorab definierte 3D-, 5D-shuffled- und Random-Graph-Kontrollen, operationalisiert als activation_auc_0_32 pro Spike und pro geliefertem Event; ein allgemeiner 5D-Vorteil wird nicht behauptet. *(untested)*
+
+**Claims:**
+- `CLAIM-S1-TOPO-001`: Unter kontrollierter Neuronenzahl, globalem Kantenbudget, Stimulus, Gewichten und Delays verändern mindestens zwei vorab definierte Netzwerktopologien die registrierten Propagationsmetriken innerhalb des definierten Stage-1-Small-SNN-Betriebsbereichs; ein Vorteil einer 5D-Anordnung wird nicht vorausgesetzt. *(inconclusive, low)* — Evidenz: EVID-2026-19
+- `CLAIM-S1-EFFICIENCY-001`: Im registrierten 64-Neuronen-Small-SNN-Regime kann eine spezifische 5D-Koordinatenorganisation nach expliziter Kontrolle dynamischer Rekrutierung eine hoehere Propagations-Effizienz als vorab definierte Kontrollen zeigen, operationalisiert durch activation_auc_0_32 pro Spike und pro geliefertem Event; dies ist kein Claim zu Speicherreduktion, Informationskapazitaet oder allgemeiner 5D-Ueberlegenheit. *(untested, none)*
 
 **Literatur:**
 - `SRC-WATTS-STROGATZ-1998`: Duncan J. Watts et al. (1998)
 - `SRC-BARABASI-1999`: Albert-László Barabási et al. (1999)
+- `SRC-OLSHAUSEN-FIELD-1996`: Bruno A. Olshausen et al. (1996)
+
+**Evidenzen:** EVID-2026-19
 
 ---
 
@@ -1693,17 +1701,22 @@
 
 **Frage:** Reagiert Brain-5D auf spike-tragende zeitliche Reihenfolge anders als auf umgekehrte oder simultane Kontrollfolgen?
 
-**Status:** open
+**Status:** inconclusive
 **Relevanz:** EXP-GEN-0021 zeigte Temporal-State-Diskrepanzen ohne Spike-Aktivität.
 
 **Aktuelle Antwort:**
 
-> EXP-S1-TEMP-ORDER-V2-20260919 liefert präregistrierte Stage-1-DATA aus 120 Läufen über 20 Seeds mit identity-destroyed Kontrolle. Der Laufstatus ist SUPPORTED_WITHIN_PREREGISTERED_PROTOCOL. Der Human Review durch Thomas Heisig ist mit accepted_as_interpretation abgeschlossen; EVID-Promotion und unabhängige Replikation stehen separat aus.
+> EXP-S1-TEMP-PROMO-R1-20260927 reproduced the bounded six-neuron two-channel temporal-order task on 20 fresh paired seeds under the current EvidenceEngine provenance contract. The canonical Human Review decision is supports with explicit perfect-score saturation and decoder/task limitations. Explicit promotion created EVID-2026-20. The historical V2 run remains reviewed DATA.
 
-*Konfidenz: preregistered_task_data_human_reviewed_evid_not_promoted*
+*Konfidenz: bounded_temporal_order_evid_with_methodological_limitations*
 
 **Hypothesen:**
-- `H-TEMP-002-A`: Forward-, Reverse- und Simultanfolgen erzeugen bei gleicher Ereignisanzahl unterscheidbare spike-basierte Antwortsignaturen. *(untested)*
+- `H-TEMP-002-A`: Forward-, Reverse- und Simultanfolgen erzeugen bei gleicher Ereignisanzahl unterscheidbare spike-basierte Antwortsignaturen. *(inconclusive)* — Evidenz: EVID-2026-20
+
+**Claims:**
+- `CLAIM-S1-TEMP-001`: Unter dem registrierten sechsneuronigen Small-SNN-Task erzeugen unterscheidbare Eingabekanäle eine decodierbare zeitliche Reihenfolge an den festgelegten Ausgängen, während die identity-destroyed Kontrolle dieses feste Ordnungssignal entfernt. *(inconclusive, low)* — Evidenz: EVID-2026-20
+
+**Evidenzen:** EVID-2026-20
 
 ---
 

research/generated/REVIEW_QUEUE.md

@@ -277,6 +277,8 @@ This projection reconciles manifest review flags with append-only review artefac
 | EXP-RETRY-AIRR-2026-0001-02 | unknown | unresolved | review_not_found_in_experiment_root |
 | EXP-S1-REC-CLEAN-R2-20260919 | PENDING | research/experiments/EXP-S1-REC-CLEAN-R2-20260919/manifest.json.review.json, research/experiments/EXP-S1-REC-CLEAN-R2-20260919/review_request.json, research/experiments/EXP-S1-REC-CLEAN-R2-20260919/review_request.json.review.json | recorded_review_present_check_status_binding |
 | EXP-S1-TEMP-ORDER-V2-20260919 | PENDING | research/experiments/EXP-S1-TEMP-ORDER-V2-20260919/manifest.json.review.json, research/experiments/EXP-S1-TEMP-ORDER-V2-20260919/review_request.json, research/experiments/EXP-S1-TEMP-ORDER-V2-20260919/review_request.json.review.json | recorded_review_present_check_status_binding |
+| EXP-S1-TEMP-PROMO-R1-20260927 | unknown | research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/human_review.json, research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/review_request.json, research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/review_request.json.review.json | recorded_review_present_check_status_binding |
+| EXP-S1-TOPO-PROMO-R1-20260927 | unknown | research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/human_review.json, research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/review_request.json, research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/review_request.json.review.json | recorded_review_present_check_status_binding |
 | EXP-S1-TOPO-V2-20260918 | unknown | research/experiments/EXP-S1-TOPO-V2-20260918/review_request.json, research/experiments/EXP-S1-TOPO-V2-20260918/review_request.json.review.json | recorded_review_present_check_status_binding |
 | EXP-S1-TOPO-V3-20260918 | unknown | unresolved | review_not_found_in_experiment_root |
 | EXP-S1-TOPO-V3-R1-20260918 | unknown | research/experiments/EXP-S1-TOPO-V3-R1-20260918/review_request.json, research/experiments/EXP-S1-TOPO-V3-R1-20260918/review_request.json.human-review.json, research/experiments/EXP-S1-TOPO-V3-R1-20260918/review_request.json.review.json | recorded_review_present_check_status_binding |

research/generated/STAGE_MATRIX.md

@@ -5,7 +5,7 @@ Stage 0-10 is a maturity/development axis, not the research-direction taxonomy.
 | Stage | Name | Recorded score | Claim boundary |
 | --- | --- | --- | --- |
 | 0 | Einzelne Nervenzelle | 0.925 | Die eingefrorene Stage-0-Konformitätsstudie stützt den numerischen Modellvertrag für Izhikevich-2003 und lif-current-v1 im deklarierten Operating Envelope. Das ist Modell-/Softwarekonformität, keine biologische Gleichwertigkeit, keine allgemeine Langzeittrajektorien-Identität und keine unabhängige Replikation. |
-| 1 | Kleines SNN | 0.75 | Die zentrale Stage-1-Topologielinie (V2 + V3-R1) und die getrennte Temporal-Order-Funktionslinie sind innerhalb ihrer präregistrierten Small-SNN-Operating-Envelopes human-reviewed DATA. Das ist weder kanonische EVID noch unabhängige Replikation und begründet keine Kognition, Skalierbarkeit oder 5D-Überlegenheit. |
+| 1 | Kleines SNN | 0.85 | Stage 1 besitzt jetzt zwei getrennte kanonische, human-reviewte EVID-Linien: Topologie EVID-2026-19 im 64-Neuronen/246-Kanten-Regime und Temporal Order EVID-2026-20 im sechsneuronigen Zweikanal-Task. Der Temporal-Human-Review dokumentiert Perfect-Score-Sättigung und die simultane Bedingung als nicht-inferentiellen Task-Adequacy-Control. Das begründet weder Kognition, Lernen/Gedächtnis, Skalierbarkeit, biologische Äquivalenz, 5D-Überlegenheit noch unabhängige Replikation. |
 | 2 | Stabiles rekurrentes SNN | 0.65 | Stabile deterministische Runs stützen Stabilität im geprüften Bereich; sie sind kein Nachweis von Gedächtnis oder Kognition. |
 | 3 | Plastisches Nervengewebe | 0.55 | STDP, Drei-Faktor-Lernen, Homöostase und Strukturplastizität sind implementierte Mechanismen; ihre wissenschaftliche Wirksamkeit bleibt interventions- und EVID-gebunden. |
 | 4 | Spezialisierte neuronale Areale | 0.55 | Modality-specific pathways und Gateway-Verträge sind technisch vorhanden; E01-E05 bleiben DATA bis zur unabhängigen wissenschaftlichen Prüfung. |

research/generated/verification/determinism_infrastructure.json

@@ -1,12 +1,12 @@
 {
   "schema_version": 1,
   "status": "verified",
-  "test_run_head": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_commit": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_tree_digest": "dcf81846537a45d30b35f70d71027eb2a58a11a80fcc2197d708bfd20c95519c",
+  "test_run_head": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_commit": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_tree_digest": "c41ed64fca8ad176f18c373e19cbf627bb92579bb54eabd610ac3afd5f5a3f25",
   "scope": "restore_determinism",
   "scope_digest": "62281a2f6208f9eabd4334c8054e2a9cca066715523823ce232846bd65cefe1a",
-  "timestamp": "2026-09-26T16:24:04+0000",
+  "timestamp": "2026-09-27T17:33:36+0000",
   "python": "3.13.15",
   "test_files": [
     "tests/test_rng_persistence.py",

research/generated/verification/phase_b_gate_status.json

@@ -3,23 +3,23 @@
     "artifact": "tests/test_baseline.json",
     "collection_errors": 0,
     "failed": 0,
-    "passed": 1677,
+    "passed": 1840,
     "python": "3.13.15",
     "skipped": 2
   },
   "continuous_integration": {
     "conclusion": "success",
-    "head_sha": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-    "run_id": 36255095320,
-    "workflow": "Verification Refresh #11"
+    "head_sha": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+    "run_id": 36337184072,
+    "workflow": "Verification Refresh #108"
   },
   "final_validation": {
     "command": "python scripts/generate_baseline.py",
     "failed": 0,
-    "passed": 1677,
+    "passed": 1840,
     "skipped": 2
   },
-  "generated": "2026-09-26T16:24:08.360708+00:00",
+  "generated": "2026-09-27T17:33:39.643959+00:00",
   "limitations": [
     "Scientific evidence status remains independent from release engineering readiness.",
     "External Hugging Face proxy smoke remains non-blocking operational follow-up."
@@ -42,8 +42,8 @@
     "overall": "passed"
   },
   "source_freeze": {
-    "commit": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-    "tested_tree_digest": "dcf81846537a45d30b35f70d71027eb2a58a11a80fcc2197d708bfd20c95519c"
+    "commit": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+    "tested_tree_digest": "c41ed64fca8ad176f18c373e19cbf627bb92579bb54eabd610ac3afd5f5a3f25"
   },
   "status": "verified",
   "verification_artifacts": [

research/generated/verification/restore_determinism.json

@@ -2,19 +2,19 @@
   "schema_version": 1,
   "suite": "restore_determinism",
   "status": "verified",
-  "test_run_head": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_commit": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_tree_digest": "dcf81846537a45d30b35f70d71027eb2a58a11a80fcc2197d708bfd20c95519c",
+  "test_run_head": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_commit": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_tree_digest": "c41ed64fca8ad176f18c373e19cbf627bb92579bb54eabd610ac3afd5f5a3f25",
   "scope": "restore_determinism",
   "scope_digest": "62281a2f6208f9eabd4334c8054e2a9cca066715523823ce232846bd65cefe1a",
-  "timestamp": "2026-09-26T16:23:14+0000",
+  "timestamp": "2026-09-27T17:32:51+0000",
   "python": "3.13.15 (main, Aug  6 2026, 02:15:18) [GCC 13.3.0]",
   "K": 500,
   "N": 1000,
   "seed": 42,
   "config_sha256": "b9dbcba2952663b8f75c839bcf935e66539f35c7ddcc1294f4157f869aae42d4",
-  "pid_C1": 12027,
-  "pid_C2": 12031,
+  "pid_C1": 11747,
+  "pid_C2": 11751,
   "digest_A": "eca7a8b2b81bc9bf3045933bd12113a31204d01355636bfd0e4432f11bed77b7",
   "digest_B": "eca7a8b2b81bc9bf3045933bd12113a31204d01355636bfd0e4432f11bed77b7",
   "digest_C": "eca7a8b2b81bc9bf3045933bd12113a31204d01355636bfd0e4432f11bed77b7",

research/generated/verification/single_listener.json

@@ -2,11 +2,11 @@
   "schema_version": 1,
   "suite": "single_listener",
   "status": "verified",
-  "timestamp": "2026-09-26T16:23:36.545997",
+  "timestamp": "2026-09-27T17:33:11.575998",
   "python_version": "3.13.15",
-  "test_run_head": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_commit": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_tree_digest": "dcf81846537a45d30b35f70d71027eb2a58a11a80fcc2197d708bfd20c95519c",
+  "test_run_head": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_commit": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_tree_digest": "c41ed64fca8ad176f18c373e19cbf627bb92579bb54eabd610ac3afd5f5a3f25",
   "scope": "runtime_integration",
   "scope_digest": "a8aa5e085d359daf039882919f138085c3987fcffb839617335b3332b5fec49f",
   "test_command": "python -m pytest tests/test_single_listener.py -q",

research/generated/verification/structural_e2e.json

@@ -2,11 +2,11 @@
   "schema_version": 1,
   "suite": "structural_e2e",
   "status": "verified",
-  "timestamp": "2026-09-26T16:23:50.376542",
+  "timestamp": "2026-09-27T17:33:24.182030",
   "python_version": "3.13.15",
-  "test_run_head": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_commit": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_tree_digest": "dcf81846537a45d30b35f70d71027eb2a58a11a80fcc2197d708bfd20c95519c",
+  "test_run_head": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_commit": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_tree_digest": "c41ed64fca8ad176f18c373e19cbf627bb92579bb54eabd610ac3afd5f5a3f25",
   "scope": "structural_e2e",
   "scope_digest": "4036baa871020d36f917b40e4e445209f6ae9f0577e5b082504fb22fe6601a28",
   "test_command": "python -m pytest tests/test_structural_e2e.py -q",

research/generated/verification/structural_live_loop.json

@@ -2,11 +2,11 @@
   "schema_version": 1,
   "suite": "structural_live_loop",
   "status": "verified",
-  "timestamp": "2026-09-26T16:23:51.621315",
+  "timestamp": "2026-09-27T17:33:25.213753",
   "python_version": "3.13.15",
-  "test_run_head": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_commit": "dcec002878a30c210d3b8ea8cfa401283b268d1c",
-  "tested_tree_digest": "dcf81846537a45d30b35f70d71027eb2a58a11a80fcc2197d708bfd20c95519c",
+  "test_run_head": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_commit": "248fd6a8ca50f2afa65453d6b87a0dc826d84f6e",
+  "tested_tree_digest": "c41ed64fca8ad176f18c373e19cbf627bb92579bb54eabd610ac3afd5f5a3f25",
   "scope": "structural_live_loop",
   "scope_digest": "ed705dc2b187a0a2deb86b1a0c4e13661e0d29f442760990b243675a450fec9a",
   "test_command": "python -m pytest tests/test_structural_live_loop.py -q",

research/notes/2026-09-27_stage1_temporal_order_decoder_followup.md

@@ -0,0 +1,45 @@
+# Methodological follow-up — Temporal-Order decoder stress test
+
+**Date:** 2026-09-27  
+**Source experiment:** `EXP-S1-TEMP-PROMO-R1-20260927`  
+**Status:** exploratory follow-up required  
+**Evidence role:** none; does not modify the present Human Review or future EVID registration
+
+## Resolved point
+
+The preregistered simultaneous condition is not part of the forward/reverse order-accuracy endpoint. It is a separate non-inferential task-adequacy control.
+
+The frozen decoder defines:
+- forward: output A first;
+- reverse: output B first;
+- simultaneous: first A and B output spikes on the same tick.
+
+The promotion DATA show simultaneous output at equal first-output ticks in the intact arm, so `intact_simultaneous_success_fraction = 1.0` is correct under the preregistered decoder. It is not a chance-level order-decoding score.
+
+## Remaining methodological concern
+
+The task saturates:
+- intact forward/reverse accuracy = 1.0;
+- identity-destroyed accuracy = 0.0;
+- simultaneous-control success = 1.0;
+- paired delta = 1.0 for all 20 seeds.
+
+This strongly supports the bounded frozen task, but provides little resolution on robustness margin or whether a simpler deterministic shortcut could also solve the task.
+
+## Prospective follow-up candidates
+
+A future preregistered stress test should consider:
+1. several nonzero order gaps, including near-simultaneous gaps;
+2. physical A/B channel-label swaps after network construction;
+3. balanced amplitude/current perturbations independent of logical order;
+4. held-out or blinded decoder mapping where feasible;
+5. increased path asymmetry controls;
+6. evaluation of whether performance degrades smoothly as the order gap approaches zero;
+7. an independently implemented decoder family (for example a separately specified threshold/readout rule) to test whether the result depends on the frozen decoder implementation;
+8. a counterfactual channel-identity condition in which A and B use identical stimulus profiles and differ only in physical channel position, to separate channel identity from amplitude/profile cues.
+
+These are prospective method-development ideas only. They are not grounds for retroactively changing the current DATA or Human Review unless a later preregistered study produces contradictory evidence.
+
+## Claim boundary
+
+The present Temporal-Order line remains restricted to a six-neuron, four-synapse, two-channel acyclic task. No learning, memory, cognition, general temporal reasoning, biological equivalence or scaling claim follows from it.

research/notes/2026-09-27_stage1_topology_efficiency_exploratory.md

@@ -0,0 +1,98 @@
+# Exploratory note — Stage-1 topology efficiency observation
+
+**Date:** 2026-09-27  
+**Source experiment:** `EXP-S1-TOPO-PROMO-R1-20260927`  
+**Authority:** exploratory / hypothesis-generating only  
+**EVID status:** none  
+**Claim:** does not modify `EVID-2026-19`
+
+## Observation
+
+The preregistered Stage-1 topology promotion run supported only the bounded topology-sensitive propagation claim `CLAIM-S1-TOPO-001`. After completion, additional ratios were calculated from the already observed condition medians. These ratios were **not preregistered endpoints** and therefore cannot be used as confirmatory evidence.
+
+From the persisted medians:
+
+| condition | activation_auc_0_32 | total_spikes | delivered_events | AUC/spike | AUC/event |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| 3d | 27.953125 | 281.0 | 1036.0 | 0.0994773 | 0.0269818 |
+| 5d | 24.828125 | 177.5 | 631.0 | 0.1398768 | 0.0393473 |
+| 5d_shuffled | 21.1171875 | 166.0 | 560.0 | 0.1272120 | 0.0377093 |
+| random_graph | 22.609375 | 171.0 | 587.0 | 0.1322186 | 0.0385168 |
+
+Relative to 3d, the median-ratio calculation is approximately:
+- 5d AUC/spike: +40.6%
+- 5d AUC/delivered-event: +45.8%
+
+These values are descriptive post-hoc observations only.
+
+## Important correction: occupancy versus dynamic recruitment
+
+The registered `5d` condition uses shape `4 x 2 x 2 x 2 x 2 = 64` and the runner constructs all 64 coordinates. Therefore the Stage-1 5d condition does **not** occupy only ~6% of its declared coordinate lattice. All 64 coordinate positions are materialized.
+
+The lower `final_active_fraction` of the 5d condition (~0.867 median) is therefore a difference in **dynamic recruitment during the run**, not a difference in structural address-space occupancy.
+
+Consequently, a prospective efficiency experiment must control dynamic recruitment/activation ceiling rather than treating address-space sparsity as an established explanation.
+
+## Competing explanations to test prospectively
+
+The post-hoc ratio pattern is compatible with several distinct explanations. None is privileged by the historical DATA:
+
+1. **Specific coordinate-organization effect.** The registered 5D graph organization itself could change how much early activation is obtained per emitted spike or delivered synaptic event.
+2. **Sparse-/efficient-coding-like activity pattern.** A lower-activity code could improve the descriptive ratios without the effect being specific to 5D geometry. This is an activity-pattern explanation, not structural address-space sparsity.
+3. **Dynamic-recruitment artifact.** Because the 5D condition recruits a smaller final fraction of neurons than 1d–3d in the historical run, a smaller denominator could mechanically increase AUC/spike or AUC/event.
+4. **Metric/window artifact.** `activation_auc_0_32` integrates activity over a fixed 32-tick window. Different onset, duration or recruitment trajectories can yield similar integrals, so a ratio based on this AUC need not identify a unique mechanism.
+
+The prospective experiment must distinguish these explanations rather than assuming that a positive ratio is evidence for 5D geometry.
+
+## Relation to the Temporal-Order line
+
+The separate Temporal-Order methodological issue around perfect decoder scores does not enter this efficiency study: the efficiency endpoints use propagation AUC, spike counts and delivered synaptic events and require **no temporal-order decoder**. The two Stage-1 lines therefore remain methodologically separate.
+
+## Sparse-coding literature boundary
+
+Olshausen & Field (1996) is included only as related-work context for the general principle that sparse activity can participate in efficient coding. The MHRN 5D coordinate construction is **not** equivalent to sparse coding in the Olshausen–Field sense, and this prospective study does not reproduce or test their natural-image coding result.
+
+## Why this is not evidence
+
+The ratios `activation_auc_0_32 / total_spikes` and `activation_auc_0_32 / delivered_events` were not preregistered primary or secondary inferential endpoints in the promotion run. They were calculated after the topology result was known.
+
+They may motivate a new hypothesis but must not be attached to `EVID-2026-19` as confirmatory support.
+
+## Prospective hypothesis generated
+
+`H-SNN-003-C` / `CLAIM-S1-EFFICIENCY-001` now capture the bounded question prospectively.
+
+The intended study must:
+1. predefine AUC/spike and AUC/event before execution;
+2. keep neuron count, edge budget, neuron model, weights/delays and stimulus accounting explicit;
+3. include controls that separate coordinate organization from dynamic-recruitment effects;
+4. use a disjoint calibration partition for any activation-matching procedure;
+5. use fresh evaluation seeds not present in the repository;
+6. freeze all endpoint, multiplicity and decision rules before execution;
+7. remain DATA until separate human review and EvidenceEngine promotion.
+
+## Claim boundary
+
+Even a positive prospective result would support only a scoped propagation-efficiency mechanism in the tested Small-SNN envelope. It would not establish memory-capacity gain, bit-level storage reduction, parameter-count reduction, Shannon information capacity, general representational capacity, cognition, biological equivalence, scaling, or universal 5D superiority. AUC/spike and AUC/event are efficiency ratios, not information-capacity measures. All registered conditions materialize 64 neurons; this study does not test materialized-coordinate count or theoretical addressability.
+
+
+## Pre-freeze calibration observation
+
+The preregistered method calibration `CAL-S1-TOPO-EFFICIENCY-R1-20260927` tested stimulus-current candidates 70–140 on disjoint calibration seeds. Across this entire registered range, the median `final_active_fraction` remained 1.0 for the 3d control and 0.828125 for the 5d control. Neither the 3d recruitment-matching gate nor the 5d high-recruitment gate passed.
+
+This is an **exploratory method-calibration observation**, not confirmatory evidence for a structural 5D mechanism. It shows only that changing the registered stimulus-current parameter over this frozen range did not erase the dynamic-recruitment difference. The cause remains unresolved and may involve graph geometry, path structure, threshold/homeostatic dynamics, or other registered model interactions.
+
+The result must not be reformulated as “5D structurally causes lower recruitment” without a new prospective study that separates those explanations. The valid gate failure blocks the R1 efficiency freeze and is retained as an auditable negative calibration result.
+
+
+## Current-sweep calibration observation
+
+The pre-freeze method calibration `CAL-S1-TOPO-EFFICIENCY-R1-20260927` produced an additional exploratory observation:
+
+- across the registered stimulus-current candidates 70–140, the 3D calibration median `final_active_fraction` remained 1.0;
+- across the same candidate range, the 5D calibration median remained 0.828125;
+- neither the preregistered 3D recruitment-matching gate nor the 5D high-recruitment gate passed.
+
+This means that **within this specific registered current range and calibration design**, changing stimulus current did not remove the observed 3D/5D recruitment difference.
+
+This observation is context only. It does not establish that the difference is a causal or universal property of 5D geometry, and it does not identify the mechanism. Network topology, path structure, coordinate organization, input/output placement and other fixed design choices remain possible explanations. No new confirmatory hypothesis or EVID is created from this calibration observation.

research/preregistrations/PREREG-S1-TEMP-PROMO-R1.json

@@ -0,0 +1,192 @@
+{
+  "schema_version": 1,
+  "preregistration_id": "PREREG-S1-TEMP-PROMO-R1",
+  "stage": 1,
+  "research_question": "RQ-TEMP-002",
+  "hypothesis": "H-TEMP-002-A",
+  "protocol": "stage1_two_channel_temporal_order_v2",
+  "mode": "CONFIRMATORY_INTERNAL_PROMOTION_REPLICATION",
+  "execution_authorized": true,
+  "scientific_evidence": false,
+  "automatic_evidence_promotion": false,
+  "human_review_required": true,
+  "independent_replication": false,
+  "rationale": "Prospective promotion-eligible internal replication of the bounded Temporal-Order Stage-1 claim. It preserves the six-neuron dual-channel task, intact and identity-destroyed arms, simultaneous control, frozen decoder and single paired inferential contrast from V2, while using fresh predeclared seeds and the current EvidenceEngine provenance contract. Historical V2 artifacts remain immutable.",
+  "task": {
+    "network": "six-neuron acyclic dual-channel SNN",
+    "paths": [
+      "input_A -> hidden_A -> output_A",
+      "input_B -> hidden_B -> output_B"
+    ],
+    "neuron_count": 6,
+    "synapse_count": 4,
+    "synaptic_delay_ticks": 1,
+    "order_gap_ticks": 4,
+    "evaluation_ticks": 16,
+    "logical_orders": {
+      "forward": [
+        [
+          0,
+          "A"
+        ],
+        [
+          4,
+          "B"
+        ]
+      ],
+      "reverse": [
+        [
+          0,
+          "B"
+        ],
+        [
+          4,
+          "A"
+        ]
+      ],
+      "simultaneous": [
+        [
+          0,
+          "A"
+        ],
+        [
+          0,
+          "B"
+        ]
+      ]
+    },
+    "decoder": "forward if first output_A spike precedes first output_B spike; reverse if B precedes A; simultaneous if their first output spikes occur on the same tick"
+  },
+  "arms": {
+    "intact": "logical A/B event identity is preserved at the two physical input channels",
+    "identity_destroyed": "both logical A/B events are injected into physical input_A; event count, event times, per-event current and total injected charge are preserved, but channel identity is deliberately removed"
+  },
+  "seed_bound_nuisance_realizations": {
+    "purpose": "prospective paired robustness realizations; not independent external replication",
+    "xor_salt": 5359309,
+    "synaptic_weight_uniform": [
+      100,
+      110
+    ],
+    "stimulus_current_uniform": [
+      100,
+      110
+    ],
+    "matching_rule": "for a given seed, the same sampled weight and current are used for all arms and orders",
+    "forbidden_use": "parameters may not be selected from observed task outcomes"
+  },
+  "evaluation": {
+    "seeds": [
+      860001,
+      860002,
+      860022,
+      860023,
+      860024,
+      860025,
+      860026,
+      860027,
+      860028,
+      860029,
+      860030,
+      860031,
+      860032,
+      860033,
+      860034,
+      860035,
+      860036,
+      860037,
+      860038,
+      860039
+    ],
+    "freshness_rule": "All promotion seeds were checked against repository text before freeze and are disjoint from the historical TEMP-002 V1 seeds 101-120, Temporal-Order V2 seeds 7101-7120, Stage-1 topology V2 seeds 2101-2120, topology V3 seeds 6101-6120, topology V3-R1 seeds 6201-6220 and topology promotion seeds 6301-6320.",
+    "primary_endpoint": "per-seed balanced order-decoding accuracy across forward and reverse trials",
+    "primary_contrast": "intact minus identity_destroyed, paired by seed",
+    "control_endpoint": "intact simultaneous-control success fraction",
+    "secondary_endpoints": [
+      "first output tick by channel",
+      "output gap error",
+      "total spikes",
+      "delivered synaptic events",
+      "state digests"
+    ]
+  },
+  "analysis": {
+    "alpha": 0.05,
+    "test": "exact two-sided paired sign test on per-seed intact-minus-identity-destroyed accuracy deltas",
+    "effect_interval": "deterministic paired bootstrap 95% CI for median accuracy delta, 5000 resamples",
+    "multiple_testing": "not applicable; exactly one primary inferential contrast",
+    "simultaneous_condition_role": "non-inferential task-adequacy control"
+  },
+  "decision_rule": {
+    "minimum_intact_order_accuracy": 0.95,
+    "maximum_identity_destroyed_order_accuracy": 0.55,
+    "minimum_intact_simultaneous_success_fraction": 0.95,
+    "minimum_paired_accuracy_delta": 0.4,
+    "supported_within_protocol": "all four directional/task gates pass, paired sign-test p < alpha, bootstrap CI excludes zero, and all design-integrity checks pass",
+    "no_predefined_support": "one or more preregistered task/inference gates fail despite intact design integrity",
+    "not_tested": "execution or design-integrity failure"
+  },
+  "integrity_constraints": [
+    "clean source tree before execution",
+    "exactly 20 fresh seed-bound paired realizations",
+    "exactly six runs per seed: two arms x three orders",
+    "six neurons and four synapses in every run",
+    "identical sampled weight/current across all six runs of a seed",
+    "identical event count and total injected charge across arms",
+    "historical temporal_order_spiking_v1 artifacts remain unchanged"
+  ],
+  "claim_boundary": "This prospective promotion replication tests only CLAIM-S1-TEMP-001 / RQ-TEMP-002 / H-TEMP-002-A in the registered six-neuron acyclic two-channel Small-SNN task. It asks whether preserved A/B channel identity supports fixed temporal-order decoding while the identity-destroyed matched control removes that signal. A successful run creates promotion-eligible DATA only; EVID still requires a separate canonical human review and explicit EvidenceEngine promotion. It does not establish learning, memory, cognition, biological equivalence, scaling, 5D superiority, general temporal reasoning or independent external replication.",
+  "review": {
+    "human_review_status": "PENDING_AFTER_EXECUTION",
+    "ai_review_does_not_satisfy_human_gate": true,
+    "automatic_promotion_forbidden": true,
+    "canonical_review_decisions": [
+      "supports",
+      "refutes",
+      "inconclusive"
+    ]
+  },
+  "status": "FROZEN_BEFORE_PROMOTION_REPLICATION",
+  "claim_id": "CLAIM-S1-TEMP-001",
+  "parent_preregistration_id": "PREREG-S1-TEMP-ORDER-V2",
+  "predecessor_experiment": "EXP-S1-TEMP-ORDER-V2-20260919",
+  "provenance": {
+    "internal_replication_only": true,
+    "independent_external_replication": false,
+    "promotion_eligible_target": true,
+    "historical_artifacts_immutable": true,
+    "fresh_seed_search_checked_before_freeze": true,
+    "parent_experiment": "EXP-S1-TEMP-ORDER-V2-20260919"
+  },
+  "evidence_engine_contract": {
+    "automatic_evidence_promotion": false,
+    "scientific_evidence_on_execution": false,
+    "require_clean_git_tree": true,
+    "require_validity_true": true,
+    "require_zero_runtime_errors": true,
+    "require_zero_fatal_errors": true,
+    "require_provenance_digests": true,
+    "require_source_freeze_sha": true,
+    "require_canonical_human_review_before_promotion": true,
+    "allowed_review_decisions": [
+      "supports",
+      "refutes",
+      "inconclusive"
+    ]
+  },
+  "local_review": {
+    "verification_command": "python scripts/verify_stage1_temporal_order_promotion_r1.py",
+    "raw_data_path": "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/data/evaluation.json",
+    "statistics_path": "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/analysis/statistics.json",
+    "report_path": "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/report.md",
+    "canonical_review_path": "research/experiments/EXP-S1-TEMP-PROMO-R1-20260927/human_review.json"
+  },
+  "frozen_on": "2026-09-27",
+  "execution_authorization": {
+    "scope": "single_execution_of_EXP-S1-TEMP-PROMO-R1-20260927",
+    "target_branch": "develop",
+    "overwrite_existing_experiment": false,
+    "automatic_evidence_promotion": false,
+    "human_review_after_execution": true
+  }
+}

research/preregistrations/PREREG-S1-TOPO-EFFICIENCY-R1.json

@@ -0,0 +1,358 @@
+{
+  "schema_version": 1,
+  "preregistration_id": "PREREG-S1-TOPO-EFFICIENCY-R1",
+  "stage": 1,
+  "research_question": "RQ-SNN-003",
+  "hypothesis": "H-SNN-003-C",
+  "claim_id": "CLAIM-S1-EFFICIENCY-001",
+  "protocol": "stage1_topology_efficiency_r1",
+  "mode": "PROSPECTIVE_CONFIRMATORY_AFTER_EXPLORATORY_OBSERVATION",
+  "status": "DRAFT_CALIBRATION_GATE_FAILED_FREEZE_BLOCKED",
+  "execution_authorized": false,
+  "scientific_evidence": false,
+  "automatic_evidence_promotion": false,
+  "human_review_required": true,
+  "independent_replication": false,
+  "exploratory_origin": {
+    "source_experiment": "EXP-S1-TOPO-PROMO-R1-20260927",
+    "source_evidence": "EVID-2026-19",
+    "note": "research/notes/2026-09-27_stage1_topology_efficiency_exploratory.md",
+    "anti_harking_rule": "The historical AUC/spike and AUC/event ratios motivate this protocol only. They are not imported as confirmatory evidence and are not part of EVID-2026-19."
+  },
+  "scientific_correction": {
+    "occupancy": "The 5d shape 4x2x2x2x2 materializes all 64 coordinates. Lower final_active_fraction is dynamic recruitment, not sparse address-space occupancy.",
+    "storage_claim": "This protocol does not test bit-level storage reduction, parameter-count reduction, memory capacity, representational capacity or general information capacity.",
+    "information_capacity": "AUC/spike and AUC/delivered-event are efficiency ratios, not Shannon-information or representational-capacity measures.",
+    "materialization": "All registered conditions materialize 64 neurons. This protocol does not test materialized-coordinate count or theoretical addressability.",
+    "related_work": "Olshausen & Field (1996) is context for sparse/efficient coding only. It does not predict that 5D geometry is efficient and is not confirmatory evidence for CLAIM-S1-EFFICIENCY-001."
+  },
+  "conditions": {
+    "3d_reference": {
+      "shape": [
+        4,
+        4,
+        4,
+        1,
+        1
+      ],
+      "graph_rule": "nearest-future labels under normalized Euclidean distance",
+      "role": "lower-dimensional geometric reference"
+    },
+    "5d_reference": {
+      "shape": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "graph_rule": "nearest-future labels under normalized Euclidean distance",
+      "role": "target geometry"
+    },
+    "5d_shuffled": {
+      "shape": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "graph_rule": "shuffle coordinate-to-label assignment before the same geometric edge construction",
+      "role": "same coordinate lattice with destroyed label/geometry alignment"
+    },
+    "random_graph": {
+      "shape": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "graph_rule": "uniform future-target sampling under the identical out-degree schedule",
+      "role": "non-geometric graph control"
+    },
+    "3d_recruitment_matched": {
+      "shape": [
+        4,
+        4,
+        4,
+        1,
+        1
+      ],
+      "graph_rule": "same as 3d_reference",
+      "role": "3d control whose stimulus-current level is chosen only on disjoint calibration seeds to match a preregistered dynamic-recruitment target"
+    },
+    "5d_high_recruitment": {
+      "shape": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "graph_rule": "same as 5d_reference",
+      "role": "mechanistic sensitivity arm only; its calibration-selected current is evaluated descriptively and is excluded from the primary Holm family and from all support/refute criteria"
+    }
+  },
+  "matched_budgets": {
+    "neuron_count": 64,
+    "out_degree_cap": 4,
+    "expected_edge_count": 246,
+    "delay_ticks": 1,
+    "synaptic_weight": 55,
+    "stimulus_input_labels": [
+      0,
+      1,
+      2,
+      3
+    ],
+    "output_labels": [
+      60,
+      61,
+      62,
+      63
+    ],
+    "stimulus_ticks": 1,
+    "evaluation_ticks": 128,
+    "auc_window_ticks": 32,
+    "note": "Reference conditions use identical neuron count, edge budget, weight, delay, stimulus accounting and evaluation window. Recruitment-control conditions may differ only in the calibration-selected stimulus-current level declared below.",
+    "readout_contract": {
+      "stimulus_procedure": "All evaluation arms stimulate physical labels 0,1,2,3 at tick 1 using the same condition-level current contract.",
+      "output_procedure": "All evaluation arms read physical output labels 60,61,62,63 using the same propagation metrics.",
+      "decoder": "none; AUC/spike and AUC/delivered-event are computed directly from registered propagation counters.",
+      "only_graph_difference": "For 3d_reference, 5d_reference, 5d_shuffled and random_graph, graph construction is the only intended structural difference. Recruitment-control arms may additionally use only their calibration-selected stimulus current."
+    }
+  },
+  "calibration": {
+    "purpose": "select recruitment-control stimulus-current levels only; no confirmatory hypothesis testing",
+    "seeds": [
+      991001,
+      991002,
+      991003,
+      991004,
+      991005,
+      991006,
+      991007,
+      991008,
+      991009,
+      991010
+    ],
+    "candidate_stimulus_currents": [
+      70,
+      80,
+      90,
+      100,
+      110,
+      120,
+      130,
+      140
+    ],
+    "reference_stimulus_current": 100,
+    "3d_recruitment_matched_selection_rule": "Choose the lowest candidate current whose calibration-seed median final_active_fraction lies in [0.84,0.90], median output_reach_fraction >=0.95 and median delivered_events >=32. If none qualifies, the recruitment-matched panel is NOT_TESTED.",
+    "5d_high_recruitment_selection_rule": "Choose the lowest candidate current whose calibration-seed median final_active_fraction >=0.98, median output_reach_fraction >=0.95 and median delivered_events >=32. If none qualifies, the high-recruitment panel is NOT_TESTED.",
+    "forbidden_selection_inputs": [
+      "AUC/spike effect size",
+      "AUC/event effect size",
+      "first-output latency effect size",
+      "p-values",
+      "evaluation-seed outcomes",
+      "direction of any 5D advantage"
+    ],
+    "persistence": {
+      "path": "research/calibrations/CAL-S1-TOPO-EFFICIENCY-R1-20260927",
+      "provenance_class": "PRE_FREEZE_METHOD_CALIBRATION",
+      "confirmatory_data": false,
+      "scientific_evidence": false,
+      "immutable_after_commit": true,
+      "evaluation_seed_access_forbidden": true,
+      "note": "Calibration outputs are persisted for auditability but are neither confirmatory DATA for CLAIM-S1-EFFICIENCY-001 nor EVID."
+    },
+    "failure_policy": {
+      "infrastructure_failure": "A run that fails before producing a verified calibration result may be retried with the identical code, seed set and candidate-current grid. The failed infrastructure attempt remains in workflow history.",
+      "valid_gate_failure": "A successfully executed and verified calibration in which a preregistered selection gate is not met is a valid negative calibration result. It may not be rerun with altered candidates under this preregistration to search for a passing value.",
+      "design_change": "If a valid calibration gate fails and a different control variable, candidate grid or selection target is scientifically justified, create a new preregistration/calibration revision with a new ID and document the reason. Do not freeze this R1 design.",
+      "freeze_rule": "Freeze is prohibited unless the required 3d_recruitment_matched calibration gate passes and its selected value is hash-bound into the frozen preregistration. The 5d_high_recruitment gate is sensitivity-only; failure is recorded and the arm becomes NOT_TESTED, but it cannot be silently retuned.",
+      "not_tested_semantics": "If the required 3d recruitment-matching gate fails, the confirmatory R1 efficiency evaluation remains NOT_TESTED and execution_authorized must stay false."
+    },
+    "observed_status_after_run": {
+      "calibration_id": "CAL-S1-TOPO-EFFICIENCY-R1-20260927",
+      "verified_execution": true,
+      "evaluation_seeds_read_or_executed": false,
+      "3d_recruitment_matched_gate_passed": false,
+      "5d_high_recruitment_gate_passed": false,
+      "freeze_blocked": true,
+      "interpretation": "Across candidate stimulus currents 70-140, calibration median final_active_fraction remained 1.0 for 3d and 0.828125 for 5d. Stimulus-current modulation in this registered range therefore did not achieve either recruitment target."
+    }
+  },
+  "evaluation": {
+    "seeds": [
+      992001,
+      992002,
+      992003,
+      992004,
+      992005,
+      992006,
+      992007,
+      992008,
+      992009,
+      992010,
+      992011,
+      992012,
+      992013,
+      992014,
+      992015,
+      992016,
+      992017,
+      992018,
+      992019,
+      992020
+    ],
+    "freshness_rule": "The exact evaluation-seed block was searched in the repository before freeze and returned no matches. It must be rechecked immediately before freeze; any collision invalidates this seed block.",
+    "primary_endpoints": [
+      {
+        "id": "activation_auc_per_spike",
+        "formula": "activation_auc_0_32 / total_spikes",
+        "zero_denominator_rule": "run is invalid if total_spikes == 0"
+      },
+      {
+        "id": "activation_auc_per_delivered_event",
+        "formula": "activation_auc_0_32 / delivered_events",
+        "zero_denominator_rule": "run is invalid if delivered_events == 0"
+      }
+    ],
+    "primary_contrasts": [
+      {
+        "left": "3d_reference",
+        "right": "5d_reference",
+        "label": "5d_vs_3d_reference"
+      },
+      {
+        "left": "5d_shuffled",
+        "right": "5d_reference",
+        "label": "5d_vs_5d_shuffled"
+      },
+      {
+        "left": "random_graph",
+        "right": "5d_reference",
+        "label": "5d_vs_random_graph"
+      },
+      {
+        "left": "3d_recruitment_matched",
+        "right": "5d_reference",
+        "label": "5d_vs_3d_recruitment_matched"
+      }
+    ],
+    "directional_difference": "right-minus-left; positive favors 5d_reference for efficiency endpoints",
+    "paired_by_seed": true,
+    "test": "exact one-sided paired sign test for predeclared positive efficiency differences",
+    "effect_interval": "deterministic paired bootstrap 95% CI for the median difference, 5000 resamples",
+    "multiple_testing": "Exactly one Holm family across exactly 8 primary tests: the four explicitly listed contrasts x the two primary efficiency endpoints. No 5d_high_recruitment contrast belongs to this family.",
+    "alpha": 0.05,
+    "latency_role": "first_output_latency_censored is exploratory/descriptive only. It is outside the primary family and cannot create, rescue or upgrade support for CLAIM-S1-EFFICIENCY-001.",
+    "sample_size": {
+      "evaluation_seeds": 20,
+      "conditions": 6,
+      "runs_per_condition_per_seed": 1,
+      "total_evaluation_runs": 120,
+      "calibration_runs": 160,
+      "calibration_breakdown": "2 calibration-only control arms x 8 candidate currents x 10 disjoint calibration seeds",
+      "total_planned_runs_if_all_calibration_candidates_execute": 280
+    },
+    "seed_selection_rationale": "992001-992020 is an arbitrary new contiguous block chosen before freeze for bookkeeping convenience only. The block was not selected from prior outcomes or effect directions and must pass the automated repository-collision gate immediately before freeze.",
+    "ratio_validity": {
+      "total_spikes_minimum_per_run": 32,
+      "delivered_events_minimum_per_run": 32,
+      "rationale": "A fixed protocol stability floor prevents near-zero denominators from dominating ratio endpoints. The threshold is frozen before evaluation and is not tuned on evaluation outcomes.",
+      "failure_semantics": "No low-denominator run is deleted. If any evaluation run violates either denominator floor, the full confirmatory experiment is classified NOT_TESTED_DENOMINATOR_VALIDITY and no primary claim decision is made.",
+      "ratio_computation": "Compute each run's ratio first; never divide condition-level aggregate numerators by aggregate denominators.",
+      "summary_statistic": "paired median difference across evaluation seeds",
+      "bootstrap": "deterministic paired bootstrap over the already computed per-seed ratios/paired differences, 5000 resamples; numerator and denominator are never bootstrapped separately",
+      "threshold_basis": "The floor 32 is a preregistered structural stability threshold equal to one half of the 64-neuron network size. It is intended to reject degenerate runs driven by only a handful of spikes/events, not to separate conditions or maximize an efficiency effect. Historical Stage-1 runs produced denominators in the hundreds, so 32 is deliberately conservative and well below the observed regime.",
+      "boundary_rule": "Exactly 32 spikes/events is valid because the rule is >=32. The ratio is computed normally at the boundary; the threshold is not a censoring or winsorization rule.",
+      "skew_note": "Ratio distributions may remain right-skewed. This is why inference uses paired per-run ratios, median differences and paired bootstrap intervals rather than mean-ratio normal approximations. The denominator floor limits numerical blow-up but does not assume symmetry.",
+      "joint_validity_rule": "A run is denominator-valid only if total_spikes >= 32 AND delivered_events >= 32. Failure of either condition invalidates the entire confirmatory evaluation under NOT_TESTED_DENOMINATOR_VALIDITY; no run is removed.",
+      "joint_validity_condition": "A run is ratio-valid only if total_spikes >= 32 AND delivered_events >= 32. Failure of either condition invalidates the confirmatory evaluation under the frozen NOT_TESTED rule; no run is removed."
+    },
+    "exploratory_endpoints": [
+      "first_output_latency_censored",
+      "final_active_fraction",
+      "total_spikes",
+      "delivered_events",
+      "activation_auc_0_32",
+      "output_reach_fraction",
+      "AUC per active neuron",
+      "injected charge"
+    ]
+  },
+  "decision_rule": {
+    "prerequisites": [
+      "design integrity passes",
+      "all 120 evaluation runs satisfy the frozen denominator-validity floor",
+      "3d_recruitment_matched calibration gate succeeds before evaluation; otherwise classification is NOT_TESTED_RECRUITMENT_CONTROL",
+      "all eight primary tests are analyzed in the single frozen Holm family"
+    ],
+    "case_A_supports_specific_effect": "SUPPORTED_WITHIN_PREREGISTERED_PROTOCOL only if 5d_reference beats 3d_reference, 5d_shuffled, random_graph and 3d_recruitment_matched on BOTH primary efficiency endpoints with positive paired-median differences, Holm-adjusted p<0.05 and paired-bootstrap CI excluding zero for all eight tests.",
+    "case_B_inconclusive_specificity": "INCONCLUSIVE_SPECIFICITY if 5d_reference beats 3d_reference on both primary efficiency endpoints but fails one or more of the shuffled/random comparisons. A narrower descriptive 5d-vs-3d efficiency difference may be reported, but CLAIM-S1-EFFICIENCY-001 is not supported.",
+    "case_B_recruitment_explanation": "REFUTES_WITHIN_PROTOCOL if the 3d, shuffled and random specificity tests all pass but 5d_reference fails to beat 3d_recruitment_matched on either primary efficiency endpoint. This is evidence against the preregistered recruitment-independent specific-efficiency claim under this protocol.",
+    "case_C_refutes": "REFUTES_WITHIN_PROTOCOL if 5d_reference fails to beat 3d_reference on either primary efficiency endpoint under the frozen Holm+CI rule.",
+    "not_tested": "NOT_TESTED if design integrity, denominator validity or the preregistered 3d recruitment-matching calibration gate fails.",
+    "high_recruitment_role": "5d_high_recruitment is descriptive mechanistic sensitivity only. It is excluded from the primary Holm family and cannot change the A/B/C decision.",
+    "equal_value_rule": "SUPPORTED, INCONCLUSIVE and REFUTES are all scientifically valid outcomes and must be persisted and reported without rescue endpoints or post-hoc reclassification."
+  },
+  "integrity_constraints": [
+    "clean source tree before execution",
+    "exactly 10 disjoint calibration seeds and 20 fresh evaluation seeds",
+    "automated seed-collision gate passes at the freeze commit",
+    "64 neurons and 246 edges in every condition",
+    "reference conditions use weight 55, delay 1 and stimulus current 100",
+    "all graph arms use the identical stimulus labels, output labels, evaluation window and propagation readout code",
+    "recruitment-control stimulus currents are selected only from calibration seeds under the frozen rules",
+    "5d_high_recruitment is descriptive sensitivity only and cannot enter the primary decision",
+    "no endpoint or contrast may be added after evaluation starts",
+    "no invalid-denominator run may be silently dropped",
+    "all null, inconclusive and adverse results are persisted",
+    "historical Stage-1 artifacts remain immutable"
+  ],
+  "claim_boundary": "This study tests only a bounded propagation-efficiency hypothesis in the registered 64-neuron Small-SNN envelope. Efficiency means activation_auc_0_32 per spike and per delivered synaptic event under frozen definitions. It does not test or establish bit-level storage reduction, parameter-count reduction, memory capacity, Shannon information capacity, general representational capacity, number of materialized coordinates, theoretical addressability, learning, cognition, biological equivalence, scaling or universal 5D superiority.",
+  "evidence_engine_contract": {
+    "execution_creates_data_only": true,
+    "automatic_evidence_promotion": false,
+    "require_clean_git_tree": true,
+    "require_validity_true": true,
+    "require_zero_runtime_errors": true,
+    "require_zero_fatal_errors": true,
+    "require_provenance_digests": true,
+    "require_source_freeze_sha": true,
+    "require_canonical_human_review_before_promotion": true
+  },
+  "review": {
+    "human_review_status": "NOT_AVAILABLE_BEFORE_EXECUTION",
+    "ai_review_does_not_satisfy_human_gate": true,
+    "automatic_promotion_forbidden": true
+  },
+  "design_sensitivity": {
+    "method": "exact one-sided paired sign test sensitivity, expressed as positive-pair probability rather than the post-hoc +40% ratio magnitude",
+    "n_pairs": 20,
+    "family_size": 8,
+    "conservative_first_holm_threshold": 0.00625,
+    "no_tie_critical_positive_pairs": 16,
+    "conservative_power_if_true_positive_pair_probability_0_80": 0.6296482639,
+    "conservative_power_if_true_positive_pair_probability_0_85": 0.8298468463,
+    "conservative_power_if_true_positive_pair_probability_0_90": 0.9568255047,
+    "interpretation": "With 20 non-tied pairs, the most stringent Holm position is crossed by at least 16 positive pairs. This is a design-sensitivity calculation, not a post-hoc power claim and not a guarantee for every endpoint because ties reduce effective n.",
+    "anti_harking": "The historical +40.6%/+45.8% descriptive ratios are not used as the assumed effect size for this calculation."
+  },
+  "seed_freeze_gate": {
+    "script": "scripts/check_stage1_efficiency_seed_freshness.py",
+    "required_before_freeze": true,
+    "scan_scope": "tracked repository files at the proposed freeze commit",
+    "allowed_occurrence": "the preregistration file itself only",
+    "collision_semantics": "Any calibration or evaluation seed occurrence in another tracked file blocks freeze until a completely new seed block is selected.",
+    "manifest_requirement": "The freeze gate must persist the scanned commit SHA, checked seed lists and collision-free result into the frozen preregistration/freeze record."
+  },
+  "execution_block": {
+    "blocked": true,
+    "reason": "Valid pre-freeze calibration completed but the required 3d_recruitment_matched gate did not pass. R1 must not be frozen or executed.",
+    "unblocking_rule": "Only a new, explicitly versioned preregistration/calibration design may address the failed control mechanism. R1 itself remains an auditable DRAFT with execution_authorized=false."
+  }
+}

research/preregistrations/PREREG-S1-TOPO-PROMO-R1.json

@@ -0,0 +1,225 @@
+{
+  "schema_version": 1,
+  "preregistration_id": "PREREG-S1-TOPO-PROMO-R1",
+  "stage": 1,
+  "research_question": "RQ-SNN-003",
+  "hypothesis": "H-SNN-003-B",
+  "protocol": "topology_propagation_v3_r1_time_resolved_replication",
+  "mode": "CONFIRMATORY_INTERNAL_PROMOTION_REPLICATION",
+  "execution_authorized": true,
+  "scientific_evidence": false,
+  "automatic_evidence_promotion": false,
+  "predecessor_experiment": "EXP-S1-TOPO-V3-R1-20260918",
+  "design_rationale": "Prospective promotion-eligible internal replication of the bounded Stage-1 topology claim. It preserves the corrected V3-R1 topology, endpoints, matched budgets and decision rule, uses fresh evaluation seeds, records the current EvidenceEngine provenance contract, and creates DATA only. Historical V2/V3/V3-R1 artifacts remain immutable.",
+  "conditions": {
+    "1d": [
+      64,
+      1,
+      1,
+      1,
+      1
+    ],
+    "2d": [
+      8,
+      8,
+      1,
+      1,
+      1
+    ],
+    "3d": [
+      4,
+      4,
+      4,
+      1,
+      1
+    ],
+    "5d": [
+      4,
+      2,
+      2,
+      2,
+      2
+    ],
+    "5d_shuffled": [
+      4,
+      2,
+      2,
+      2,
+      2
+    ],
+    "random_graph": [
+      4,
+      2,
+      2,
+      2,
+      2
+    ]
+  },
+  "matched_budgets": {
+    "neuron_count": 64,
+    "out_degree_cap": 4,
+    "edge_count_rule": "sum_i min(4, N-i-1), identical for all conditions",
+    "expected_edge_count": 246,
+    "delay_ticks": 1,
+    "stimulus_input_labels": [
+      0,
+      1,
+      2,
+      3
+    ],
+    "output_labels": [
+      60,
+      61,
+      62,
+      63
+    ],
+    "stimulus_current": 100,
+    "stimulus_ticks": 1,
+    "evaluation_ticks": 128,
+    "synaptic_weight": 55
+  },
+  "provenance": {
+    "fixed_weight_source": "EXP-S1-TOPO-V2-20260918 preregistered activity calibration selected 55.0 before V2 evaluation",
+    "internal_replication_only": true,
+    "independent_external_replication": false,
+    "promotion_eligible_target": true,
+    "seed_overlap_forbidden_with_historical_topology_runs": true,
+    "historical_seed_ranges": [
+      "2101-2120",
+      "6101-6120",
+      "6201-6220"
+    ],
+    "analysis_contract": "Single Holm correction across all 10 time-resolved primary tests; Holm correction across the 5 first-output-latency replication contrasts.",
+    "predecessor_data_line": [
+      "EXP-S1-TOPO-V2-20260918",
+      "EXP-S1-TOPO-V3-R1-20260918"
+    ]
+  },
+  "evaluation": {
+    "seeds": [
+      6301,
+      6302,
+      6303,
+      6304,
+      6305,
+      6306,
+      6307,
+      6308,
+      6309,
+      6310,
+      6311,
+      6312,
+      6313,
+      6314,
+      6315,
+      6316,
+      6317,
+      6318,
+      6319,
+      6320
+    ],
+    "primary_endpoints": [
+      "activation_auc_0_32",
+      "half_activation_latency_censored"
+    ],
+    "replication_endpoint": "first_output_latency_censored",
+    "descriptive_endpoints": [
+      "final_active_fraction",
+      "output_reach_fraction",
+      "total_spikes",
+      "delivered_events",
+      "last_activity_tick"
+    ],
+    "primary_contrasts": [
+      [
+        "1d",
+        "2d"
+      ],
+      [
+        "2d",
+        "3d"
+      ],
+      [
+        "3d",
+        "5d"
+      ],
+      [
+        "5d",
+        "5d_shuffled"
+      ],
+      [
+        "5d",
+        "random_graph"
+      ]
+    ],
+    "low_dimensional_ceiling_resolution_contrasts": [
+      [
+        "1d",
+        "2d"
+      ],
+      [
+        "2d",
+        "3d"
+      ]
+    ],
+    "paired_by_seed": true,
+    "primary_test": "exact two-sided paired sign test on non-zero differences",
+    "primary_multiple_testing": "Single Holm correction across all 10 primary time-resolved endpoint/contrast tests",
+    "replication_test": "exact two-sided paired sign test on non-zero first-output-latency differences",
+    "replication_multiple_testing": "Holm correction across the 5 replication contrasts",
+    "effect_interval": "deterministic paired bootstrap 95% CI for the median difference, 5000 resamples",
+    "alpha": 0.05,
+    "auc_window_ticks": 32,
+    "half_activation_threshold": 0.5,
+    "censor_sentinel": 129,
+    "endpoint_definitions": {
+      "activation_auc_0_32": "Sum over ticks 0..31 inclusive of cumulative_unique_active_neurons(t)/64, evaluated after each network.step(). This is a discrete cumulative-activation area measure, not a continuous-time integral.",
+      "half_activation_latency_censored": "First zero-based tick at which cumulative_unique_active_neurons(t)/64 >= 0.5; if never reached within the 128-tick evaluation window, encode 129.",
+      "first_output_latency_censored": "First zero-based tick with any spike in output labels 60..63; if absent within 128 ticks, encode 129."
+    }
+  },
+  "decision_rule": {
+    "ceiling_resolution_supported": "At least one preregistered low-dimensional contrast (1d->2d or 2d->3d) is significant after Holm correction for activation_auc_0_32 or half_activation_latency_censored and its bootstrap CI excludes zero.",
+    "replication_supported": "All five first_output_latency_censored contrasts have the same negative right-minus-left direction observed in V2, Holm-adjusted p < 0.05, and bootstrap CI excluding zero.",
+    "supported_within_protocol": "design integrity PASS AND the single-family Holm analysis across all 10 primary tests yields ceiling_resolution_supported AND replication_supported.",
+    "partial_support": "design integrity PASS and exactly one of ceiling_resolution_supported or replication_supported is true.",
+    "no_detected_support": "design integrity PASS and neither criterion is met.",
+    "not_tested": "execution or integrity failure occurs."
+  },
+  "interpretation_boundary": "This prospective promotion replication tests only CLAIM-S1-TOPO-001 / RQ-SNN-003 / H-SNN-003-B in the registered 64-neuron, 246-edge Small-SNN operating envelope. A successful run creates promotion-eligible DATA only; EVID still requires a separate canonical human review and explicit EvidenceEngine promotion. It is not independent external replication and does not establish 5D superiority, scaling, cognition or biological equivalence.",
+  "local_review": {
+    "verification_command": "python scripts/verify_stage1_topology_promotion_r1.py",
+    "raw_data_paths": [
+      "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/data/evaluation.json"
+    ],
+    "statistics_path": "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/analysis/statistics.json",
+    "report_path": "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/report.md",
+    "canonical_review_path": "research/experiments/EXP-S1-TOPO-PROMO-R1-20260927/human_review.json"
+  },
+  "status": "FROZEN_BEFORE_PROMOTION_REPLICATION",
+  "claim_id": "CLAIM-S1-TOPO-001",
+  "evidence_engine_contract": {
+    "automatic_evidence_promotion": false,
+    "scientific_evidence_on_execution": false,
+    "require_clean_git_tree": true,
+    "require_validity_true": true,
+    "require_zero_runtime_errors": true,
+    "require_zero_fatal_errors": true,
+    "require_provenance_digests": true,
+    "require_source_freeze_sha": true,
+    "require_canonical_human_review_before_promotion": true,
+    "allowed_review_decisions": [
+      "supports",
+      "refutes",
+      "inconclusive"
+    ]
+  },
+  "frozen_on": "2026-09-27",
+  "execution_authorization": {
+    "scope": "single_execution_of_EXP-S1-TOPO-PROMO-R1-20260927",
+    "target_branch": "develop",
+    "overwrite_existing_experiment": false,
+    "automatic_evidence_promotion": false,
+    "human_review_after_execution": true
+  }
+}

research/preregistrations/PREREG-S1-TOPO-REFERENCE-R1.freeze.json

@@ -0,0 +1,82 @@
+{
+  "canonical_target_digest": "f3616151a7ab529c610f983f4fe855b84228505d806d516c502014a576d6d3b6",
+  "decision_rule_digest": "2eca23d888f4846488a2a2db53b1702120580283c5d5a73dae26667c854a01b0",
+  "evaluation_seeds": [
+    810000001,
+    810000002,
+    810000003,
+    810000004,
+    810000005,
+    810000006,
+    810000007,
+    810000008,
+    810000009,
+    810000010,
+    810000011,
+    810000012,
+    810000013,
+    810000014,
+    810000015,
+    810000016,
+    810000017,
+    810000018,
+    810000019,
+    810000020
+  ],
+  "execution_authorized": false,
+  "freeze_commit": "7c5fd5a8fa9addf058f462375a1b022a3d193a45",
+  "freeze_date": "2026-09-27",
+  "integrator_parity": {
+    "brian2_version": "2.10.1",
+    "multi_tick_contains_spike": true,
+    "multi_tick_pass": true,
+    "pass": true,
+    "path": "research/calibrations/CAL-S1-TOPO-REFERENCE-INTEGRATOR-R1/result.json",
+    "sha256": "4779cdd5e4c8993bdc4f24a30a2d6c9206cd57f7606c107aa8dc2ea13388ea05",
+    "single_tick_pass": true,
+    "tolerance_abs": 1e-12
+  },
+  "mechanism_audit": {
+    "pass": true,
+    "path": "research/audits/STAGE1_TOPOLOGY_REFERENCE_MECHANISM_AUDIT_20260927.json",
+    "sha256": "2269d45cd8d40bbcf585c8f1de2c20072273ffbcd5fe2cc028dc8a540912eef4"
+  },
+  "note": "This record freezes the scientific comparison contract only. The full Brian2 reference runner must still be implemented, independently scanned, hash-bound and separately authorized before any evaluation DATA.",
+  "preregistration_id": "PREREG-S1-TOPO-REFERENCE-R1",
+  "preregistration_sha256_before_status_update": "39b3932be79470d713023b7f56f211a32c68d15439cfdb6672802c59ea962d8f",
+  "reference_data_created": false,
+  "reference_runner": {
+    "execution_authorized": false,
+    "path": "reference/stage1_topology_brian2/runner.py",
+    "sha256": "1cbf915f86aaccac74e5c1a8c24952f132ea10d5a8b081fc5b058f24c8677ae8"
+  },
+  "reference_verifier": {
+    "path": "scripts/verify_stage1_topology_reference_r1.py",
+    "sha256": "835596d543a6f2bc3a4f7a8638fa2b310afaf1bdd5e7d0dfd7ed9210f08f249a"
+  },
+  "replication_credit_awarded": false,
+  "reset_parity": {
+    "pass": true,
+    "path": "research/calibrations/CAL-S1-TOPO-REFERENCE-RESET-R1/result.json",
+    "sha256": "c1c50e07345e940b0cfe0599cabbbbd8b2b6f0447841fa989673ddfce522b247",
+    "tolerance_abs": 1e-12
+  },
+  "sanitized_runner_protocol": {
+    "path": "reference/stage1_topology_brian2/reference_protocol.json",
+    "sha256": "9dcdef1fc9126afcf233d7d360b3fc719f2fd6108dbd8d0636df3e0dc18e0ab4"
+  },
+  "schema_version": 1,
+  "scientific_evidence": false,
+  "synapse_delay_parity": {
+    "delay_ticks": 1,
+    "pass": true,
+    "path": "research/calibrations/CAL-S1-TOPO-REFERENCE-SYNAPSE-R1/result.json",
+    "sha256": "42b051286007b315c5400871d893883d34be2847d5f165b08a70a91d0580907e",
+    "tolerance_abs": 1e-12,
+    "weight": 55.0
+  },
+  "translation_specification": {
+    "path": "reference/stage1_topology_brian2/TRANSLATION.md",
+    "sha256": "ebcb5bf58ee9fb48e9d866f95c93697ec848e5c1b1048343ea255af05ec92b2c"
+  }
+}

research/preregistrations/PREREG-S1-TOPO-REFERENCE-R1.json

@@ -0,0 +1,521 @@
+{
+  "schema_version": 1,
+  "preregistration_id": "PREREG-S1-TOPO-REFERENCE-R1",
+  "stage": 1,
+  "research_question": "RQ-SNN-003",
+  "hypothesis": "H-SNN-003-B",
+  "claim_id": "CLAIM-S1-TOPO-001",
+  "protocol": "stage1_topology_cross_implementation_reference_r1",
+  "mode": "PROSPECTIVE_CROSS_IMPLEMENTATION_PARTIAL_REPLICATION",
+  "status": "FROZEN_BEFORE_REFERENCE_RUNNER",
+  "execution_authorized": false,
+  "scientific_evidence": false,
+  "automatic_evidence_promotion": false,
+  "replication_credit_target": "partial_only",
+  "maturity_target_if_successful": {
+    "independent_replication_component": "partial",
+    "contribution": 0.075,
+    "stage1_maturity": "85% -> 92.5%",
+    "full_external_replication_still_required_for_100_percent": true
+  },
+  "independence_contract": {
+    "framework": "Brian2",
+    "framework_version": "2.10.1",
+    "python_version": "3.13",
+    "implementation_location": "reference/stage1_topology_brian2/",
+    "separate_package_required": true,
+    "forbidden_import_prefixes": [
+      "src",
+      "scripts.run_stage1_topology",
+      "research"
+    ],
+    "forbidden_runtime_dependencies": [
+      "MHRN NeuralNetwork",
+      "MHRN Neuron",
+      "MHRN Synapse",
+      "MHRN topology builders",
+      "MHRN experiment runner modules"
+    ],
+    "allowed_shared_inputs": [
+      "frozen numeric neuron/synapse/stimulus translation contract from this preregistration",
+      "frozen topology shapes and graph rules described textually in this preregistration",
+      "evaluation seed list"
+    ],
+    "runner_blinding": "The reference runner must not contain or load the canonical MHRN effect sizes, equivalence bounds, EVID files or historical experiment statistics. Replication classification is performed only by a separate verifier after raw reference DATA are persisted.",
+    "authorship_limit": "The implementation remains under the same project leadership and therefore can earn only partial cross-implementation replication credit, never full independent external replication.",
+    "implementer": "same MHRN project leadership unless an external implementation is explicitly documented later",
+    "credit_semantics": "Because the current implementation is project-side, success can only be PARTIAL_REPLICATION. A later genuinely external implementation must use a separately registered protocol and cannot inherit full-credit status from this run.",
+    "automated_leakage_guard": "A pre-freeze AST/text scan must reject imports from MHRN runtime modules and references to EVID-2026-19, EXP-S1-TOPO-PROMO-R1-20260927, canonical_targets, frozen_bounds, equivalence bounds, research/experiments, or research/registry/evidence from within reference/stage1_topology_brian2/.",
+    "runner_source": "reference/stage1_topology_brian2/runner.py",
+    "verifier_source": "scripts/verify_stage1_topology_reference_r1.py",
+    "execution_guard": "The Brian2 runner refuses execution unless MHRN_STAGE1_REFERENCE_EXECUTION=AUTHORIZED_REFERENCE_R1 is explicitly supplied by a later approved execution workflow."
+  },
+  "source_target": {
+    "canonical_experiment": "EXP-S1-TOPO-PROMO-R1-20260927",
+    "canonical_evidence": "EVID-2026-19",
+    "canonical_operating_envelope": {
+      "neurons": 64,
+      "edges": 246,
+      "dt_ms": 1,
+      "delay_ticks": 1,
+      "synaptic_weight": 55,
+      "stimulus_current": 100,
+      "stimulus_ticks": 1,
+      "evaluation_ticks": 128,
+      "auc_window_ticks": 32,
+      "input_labels": [
+        0,
+        1,
+        2,
+        3
+      ],
+      "output_labels": [
+        60,
+        61,
+        62,
+        63
+      ]
+    },
+    "conditions": {
+      "1d": [
+        64,
+        1,
+        1,
+        1,
+        1
+      ],
+      "2d": [
+        8,
+        8,
+        1,
+        1,
+        1
+      ],
+      "3d": [
+        4,
+        4,
+        4,
+        1,
+        1
+      ],
+      "5d": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "5d_shuffled": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "random_graph": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ]
+    }
+  },
+  "translation_contract": {
+    "neuron_model": "Izhikevich 2003 regular-spiking state variables v,u",
+    "initial_state": {
+      "v": -65,
+      "u": -13
+    },
+    "parameters": {
+      "a": 0.02,
+      "b": 0.2,
+      "c": -65,
+      "d": 8,
+      "threshold_mV": 30
+    },
+    "integration": "At each 1 ms tick, use the MHRN canonical two-half-Euler update for v followed by one Euler update for u: v += 0.5*(0.04*v^2+5*v+140-u+I), repeated twice; u += a*(b*v-u). A spike is detected after integration when v>=30, then reset v=c and u=u+d.",
+    "adaptation_and_secondary_mechanisms": {
+      "threshold_adaptation": {
+        "status": "ACTIVE_AND_SPIKE_TIMING_RELEVANT",
+        "parameters": {
+          "rate": 0.01,
+          "decay": 0.999,
+          "base_izhikevich_threshold": 30
+        },
+        "rule": "Spike detection uses 30.0 + threshold_adaptation. On spike, threshold_adaptation += 0.01; after firing-rate update it is multiplied by 0.999; homeostasis is then applied.",
+        "reference_requirement": "must be translated exactly"
+      },
+      "homeostasis": {
+        "status": "ACTIVE_AND_SPIKE_TIMING_RELEVANT",
+        "parameters": {
+          "target_rate_hz": 10,
+          "firing_rate_tau_ms": 1000,
+          "learning_rate": 0.001
+        },
+        "rule": "After each tick, firing_rate_estimate is updated with alpha=1-exp(-elapsed_ms/1000); homeostasis then adds 0.001*(firing_rate_estimate-10) to threshold_adaptation and clamps it to [-10,10].",
+        "reference_requirement": "must be translated exactly"
+      },
+      "energy": {
+        "status": "ACTIVE_STATE_BUT_SPIKE_TIMING_INERT_IN_CANONICAL_CORE",
+        "parameters": {
+          "initial": 1,
+          "spike_cost": 0.001,
+          "recovery_rate": 0.0001
+        },
+        "rule": "Energy is decremented on spikes and recovers after each tick, but Neuron.step/integrate_membrane/spike_threshold do not read energy to determine v/u or spike threshold.",
+        "reference_requirement": "may be omitted from the reference dynamics only if the pre-freeze source audit remains green; omission must be declared in provenance"
+      },
+      "traces": {
+        "status": "ACTIVE_STATE_BUT_SPIKE_TIMING_INERT_IN_CANONICAL_CORE",
+        "parameters": {
+          "decay": 0.95,
+          "increment": 1
+        },
+        "rule": "Pre/post traces increment/decay but are not read by membrane integration or spike threshold in the canonical topology runner.",
+        "reference_requirement": "may be omitted from the reference dynamics only if the pre-freeze source audit remains green"
+      },
+      "synaptic_plasticity": {
+        "status": "NOT_APPLIED_BY_CANONICAL_NETWORK_STEP",
+        "rule": "The topology runner creates fixed-weight Synapse objects. NeuralNetwork.step queues and delivers connection.weight but does not invoke pair-STDP, eligibility or reward updates.",
+        "reference_requirement": "reference synaptic weights remain fixed during each run"
+      },
+      "refractory": {
+        "status": "INACTIVE",
+        "refractory_ticks": 0,
+        "reference_requirement": "no additional refractory period"
+      }
+    },
+    "synapses": "Current-based excitatory events of fixed weight 55.0 with exactly one-tick delay. No plasticity during the run.",
+    "tick_order": "For tick t: clear/read external currents; deliver events whose delivery_tick==t; combine external+synaptic current; execute two half-Euler v updates and one u Euler update; test v >= 30 + current threshold_adaptation; if spiked reset v=c,u+=d and add threshold_adaptation_rate; update firing-rate estimate; decay threshold adaptation; apply homeostasis; queue outgoing fixed-weight events for t+delay. Energy/traces may update as audited but are spike-timing inert.",
+    "stimulus": "At tick 0 only, inject current 100.0 into logical labels 0,1,2,3.",
+    "topology_generation": {
+      "coordinate_order": "Generate the full lexicographic Cartesian product of the declared 5D shape, then sort coordinates by (sum(value_i/max(size_i-1,1)), coordinate_tuple), exactly matching the canonical label-order semantics while remaining independently implemented.",
+      "future_only_edges": true,
+      "out_degree_cap": 4,
+      "edge_budget": 246,
+      "geometric_rule": "For each source label, connect to the up to four future labels with smallest normalized Euclidean coordinate distance; ties are resolved by a frozen seed-derived deterministic tie key, then label.",
+      "5d_shuffled": "Shuffle coordinate-to-label assignment using the frozen reference implementation's independently specified deterministic PRNG mapping before applying the same geometric edge rule.",
+      "random_graph": "For each source, choose the same number of future targets uniformly without replacement using the frozen reference PRNG mapping.",
+      "self_connections": false,
+      "parallel_connections": false,
+      "coordinate_order_correction": {
+        "date": "2026-09-27",
+        "status": "PRE_FREEZE_CORRECTION_BEFORE_REFERENCE_DATA",
+        "previous_wording": "canonical lexicographic product order",
+        "reason": "Code-backed audit of scripts/run_stage1_topology_v2.py showed that canonical_coords sorts the Cartesian product by normalized coordinate score and then coordinate tuple; pure lexicographic wording was incomplete.",
+        "data_seen": false
+      }
+    },
+    "metric_translation": {
+      "active_neuron": "a neuron that emitted at least one spike by the end of the evaluated tick",
+      "final_active_fraction": "number of ever-active neurons divided by 64",
+      "activation_auc_0_32": "sum of cumulative ever-active-neuron fraction over ticks 0..31 inclusive",
+      "first_output_latency_censored": "first tick with any spike from labels 60..63; if absent use the frozen censor sentinel 129",
+      "delivered_events": "count of synaptic spike events delivered to postsynaptic neurons",
+      "total_spikes": "total emitted spikes across all neurons"
+    },
+    "deviation_policy": "Any discovered semantic mismatch that could change spike timing, event delivery or registered endpoints blocks execution and requires a new preregistration revision before reference DATA are generated."
+  },
+  "evaluation": {
+    "seeds": [
+      810000001,
+      810000002,
+      810000003,
+      810000004,
+      810000005,
+      810000006,
+      810000007,
+      810000008,
+      810000009,
+      810000010,
+      810000011,
+      810000012,
+      810000013,
+      810000014,
+      810000015,
+      810000016,
+      810000017,
+      810000018,
+      810000019,
+      810000020
+    ],
+    "seed_selection_rationale": "The originally drafted block 1200021-1200040 failed the executable freshness gate because seed 1200035 already existed in historical Batch artifacts. Before any reference DATA and before freeze, the entire block was discarded. The replacement block 810000001-810000020 was selected as a new contiguous bookkeeping block and returned no repository-search matches at selection time; the executable freshness gate must still pass at the exact freeze commit.",
+    "conditions": [
+      "1d",
+      "2d",
+      "3d",
+      "5d",
+      "5d_shuffled",
+      "random_graph"
+    ],
+    "total_runs": 120,
+    "primary_endpoint": "first_output_latency_censored",
+    "secondary_endpoint": "activation_auc_0_32",
+    "primary_contrasts": [
+      [
+        "1d",
+        "2d"
+      ],
+      [
+        "2d",
+        "3d"
+      ],
+      [
+        "3d",
+        "5d"
+      ],
+      [
+        "5d",
+        "5d_shuffled"
+      ],
+      [
+        "5d",
+        "random_graph"
+      ]
+    ],
+    "primary_difference": "right-minus-left",
+    "primary_test": "exact two-sided paired sign test within reference DATA",
+    "primary_multiple_testing": "one Holm family across the five first-output-latency contrasts",
+    "alpha": 0.05,
+    "effect_interval": "deterministic paired bootstrap 95% CI for median difference",
+    "secondary_role": "activation_auc_0_32 direction is a secondary concordance check and cannot rescue failed primary replication criteria",
+    "seed_revision_history": [
+      {
+        "date": "2026-09-27",
+        "old_block": "1200021-1200040",
+        "reason": "Executable pre-freeze freshness gate found historical collision for 1200035 in EXP-BATCH-20260914074039-77 artifacts.",
+        "action": "Discarded the entire block before freeze and before any reference evaluation; no outcomes existed or were inspected.",
+        "new_block": "810000001-810000020"
+      }
+    ]
+  },
+  "canonical_targets": {
+    "source": "EXP-S1-TOPO-PROMO-R1-20260927 analysis/statistics.json",
+    "first_output_latency_median_deltas": {
+      "1d_to_2d": -10,
+      "2d_to_3d": -3,
+      "3d_to_5d": -1,
+      "5d_to_5d_shuffled": -3,
+      "5d_to_random_graph": -2.5
+    },
+    "equivalence_bounds_rule": "Strict sign-preserving equivalence fixed at 50%-150% of the absolute canonical median effect for every primary latency contrast. No resolution-based widening is used.",
+    "frozen_bounds": {
+      "1d_to_2d": [
+        -15,
+        -5
+      ],
+      "2d_to_3d": [
+        -4.5,
+        -1.5
+      ],
+      "3d_to_5d": [
+        -1.5,
+        -0.5
+      ],
+      "5d_to_5d_shuffled": [
+        -4.5,
+        -1.5
+      ],
+      "5d_to_random_graph": [
+        -3.75,
+        -1.25
+      ]
+    },
+    "activation_auc_direction_only": {
+      "1d_to_2d": "positive",
+      "2d_to_3d": "positive",
+      "3d_to_5d": "negative",
+      "5d_to_5d_shuffled": "negative",
+      "5d_to_random_graph": "negative"
+    },
+    "direction_rule": "Each primary contrast must separately retain the canonical non-zero sign and its median must lie inside the frozen strict 50%-150% equivalence interval. For 3d->5d this means median <0 and within [-1.5,-0.5].",
+    "resolution_note": "The primary latency metric is discrete and paired medians with n=20 can occur in 0.5-tick increments. This discreteness is treated as a known sensitivity limitation, not as grounds to widen the equivalence interval. A Brian2 median of -2.0 for 3d->5d is outside the preregistered equivalence bound.",
+    "other_bounds_unchanged": true,
+    "criterion_revision_history": [
+      {
+        "date": "2026-09-27",
+        "status": "DRAFT_ONLY_WITHDRAWN_BEFORE_REFERENCE_DATA",
+        "change": "A temporary DRAFT considered widening 3d->5d from [-1.5,-0.5] to [-2.0,-0.5] for metric discreteness.",
+        "resolution": "Withdrawn after methodological review before any Brian2 reference evaluation DATA existed. Canonical freeze candidate restored the original strict 50%-150% interval [-1.5,-0.5].",
+        "data_seen": false
+      }
+    ]
+  },
+  "decision_rule": {
+    "partial_replication": [
+      "reference design integrity passes",
+      "all five first-output-latency reference contrasts retain the canonical direction",
+      "all five reference median deltas lie inside their frozen equivalence bounds",
+      "all five primary paired sign tests remain Holm-significant with bootstrap CI excluding zero",
+      "at least four of five activation_auc_0_32 secondary contrast directions agree with the canonical direction"
+    ],
+    "failed_replication": [
+      "any primary latency contrast reverses canonical direction",
+      "or two or more primary latency contrasts have median effects outside the frozen equivalence bounds and their 95% bootstrap CIs are wholly outside those bounds",
+      "or fewer than three of five primary latency contrasts are Holm-significant"
+    ],
+    "inconclusive": [
+      "a semantic translation mismatch is discovered after freeze",
+      "or one or more primary latency CIs span an equivalence boundary such that success/failure classification is unstable",
+      "or exactly three or four primary contrasts satisfy both Holm significance and equivalence while none reverse direction",
+      "or exactly one primary contrast is wholly outside its equivalence bound without direction reversal while the remaining four satisfy the primary criteria",
+      "or reference runtime behavior makes a preregistered endpoint non-comparable"
+    ],
+    "equal_value_rule": "PARTIAL_REPLICATION, FAILED_REPLICATION and INCONCLUSIVE_REFERENCE_REPLICATION are equally reportable scientific outcomes. No secondary endpoint or post-hoc tuning may rescue a failed primary classification."
+  },
+  "provenance_requirements": {
+    "separate_code_digest": true,
+    "framework_version_recorded": true,
+    "python_version_recorded": true,
+    "dependency_lock_recorded": true,
+    "clean_git_tree": true,
+    "runtime_and_platform_recorded": true,
+    "raw_reference_data_persisted": true,
+    "runner_source_hash_persisted": true,
+    "verifier_source_hash_persisted": true,
+    "no_shared_mhrn_code_imports_verified": true
+  },
+  "maturity_semantics": {
+    "success": "A successful reference replication counts only as partial independent/reference replication = 7.5/15 points, advancing Stage-1 maturity from 85% to 92.5%.",
+    "failure": "A failed or inconclusive reference replication does not increase maturity and must remain visible in the canonical Stage-1 state.",
+    "full_credit": "15/15 replication credit remains reserved for genuinely external, independently controlled/authored replication."
+  },
+  "review": {
+    "human_review_required": true,
+    "automatic_evidence_promotion": false,
+    "execution_creates_reference_data_only": true
+  },
+  "prefreeze_requirements": {
+    "integrator_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-INTEGRATOR-R1/result.json",
+      "protocol": "one neuron, one tick, current=0 and current=100; compare MHRN and Brian2 final v,u, threshold_adaptation, firing_rate_estimate, spike flag",
+      "tolerance_abs": 1e-12,
+      "pass_rule": "all compared scalar fields match within 1e-12 and spike flags are identical in both cases",
+      "failure_rule": "freeze blocked; no reference evaluation may run"
+    },
+    "mechanism_audit": {
+      "required": true,
+      "artifact": "research/audits/STAGE1_TOPOLOGY_REFERENCE_MECHANISM_AUDIT_20260927.json",
+      "pass_rule": "every spike-timing-relevant canonical mechanism is classified and translated; every omitted mechanism is source-demonstrated spike-timing inert",
+      "failure_rule": "freeze blocked"
+    },
+    "runner_independence_scan": {
+      "required": true,
+      "script": "scripts/check_stage1_reference_independence.py",
+      "pass_rule": "no forbidden MHRN imports, EVID/experiment access, canonical target/equivalence access, or research artifact reads in the reference runner package",
+      "failure_rule": "freeze blocked"
+    },
+    "seed_freshness": {
+      "required": true,
+      "script": "scripts/check_stage1_reference_seed_freshness.py",
+      "pass_rule": "All evaluation seeds must be absent from prior experiment/DATA use. Allowed tracked declarations are limited to the canonical preregistration, sanitized Brian2 runner protocol, the explicitly invalidated premature-freeze audit (which records reference_data_created=false), and after a valid freeze the canonical freeze record.",
+      "failure_rule": "select a new seed block before freeze",
+      "allowed_seed_declarations": [
+        "research/preregistrations/PREREG-S1-TOPO-REFERENCE-R1.json",
+        "reference/stage1_topology_brian2/reference_protocol.json",
+        "research/audits/invalidated/PREREG-S1-TOPO-REFERENCE-R1.freeze.68cc945.invalidated.json"
+      ]
+    },
+    "reset_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-RESET-R1/result.json",
+      "protocol": "one neuron, forced spike followed by the next tick; compare spike flag, reset v=c, u increment by d, threshold adaptation, firing-rate estimate and next-tick state",
+      "tolerance_abs": 1e-12,
+      "pass_rule": "all compared state fields and spike decisions match MHRN within 1e-12",
+      "failure_rule": "freeze blocked"
+    },
+    "synapse_delay_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-SYNAPSE-R1/result.json",
+      "protocol": "two neurons, one fixed excitatory synapse weight 55 and delay 1 tick; induce a source spike at tick 0 and compare event delivery tick, delivered-event count, postsynaptic current/state and spike decision",
+      "tolerance_abs": 1e-12,
+      "pass_rule": "event is absent at source-spike tick, delivered exactly one tick later, and compared target state/spike fields match MHRN within 1e-12",
+      "failure_rule": "freeze blocked"
+    }
+  },
+  "prefreeze_review_record": {
+    "review_date": "2026-09-27",
+    "status": "REVIEW_REQUIREMENTS_INTEGRATED_AWAITING_EXECUTABLE_GATES",
+    "integrated_requirements": [
+      "single- and multi-tick integrator parity before freeze",
+      "code-backed adaptation/homeostasis/energy/traces audit before freeze",
+      "strict 50%-150% sign-preserving equivalence bounds; no resolution-based widening",
+      "explicit reset parity",
+      "explicit two-neuron synapse/delay parity",
+      "automated runner leakage/independence scan",
+      "seed freshness at exact freeze commit",
+      "complete blinded Brian2 runner and separate verifier implemented before freeze",
+      "complete PARTIAL_REPLICATION / FAILED_REPLICATION / INCONCLUSIVE_REFERENCE_REPLICATION decision rules"
+    ],
+    "criterion_change_history": [
+      "Temporary DRAFT widening of 3d->5d to [-2.0,-0.5] was withdrawn before reference DATA; strict [-1.5,-0.5] restored.",
+      "Earlier canonical freeze attempt at commit 68cc945 was invalidated because later mandatory pre-freeze gates were not yet integrated."
+    ],
+    "reference_data_seen": false
+  },
+  "pre_freeze_gates": {
+    "integrator_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-INTEGRATOR-R1/result.json",
+      "rule": "PASS required before freeze. Single-tick subthreshold/spike cases and persistent multi-tick trajectory must match v/u/adaptation/firing-rate state at absolute tolerance 1e-12; the multi-tick trajectory must include at least one spike.",
+      "status": "PENDING"
+    },
+    "mechanism_audit": {
+      "required": true,
+      "artifact": "research/audits/2026-09-27_stage1_reference_mechanism_audit.md",
+      "rule": "Threshold adaptation and homeostasis are timing-relevant and must be translated. Energy and traces may be omitted only as explicitly audited causally inert state for registered propagation endpoints.",
+      "status": "PASS_DOCUMENTED"
+    },
+    "independence_scan": {
+      "required": true,
+      "script": "scripts/check_stage1_reference_independence.py",
+      "rule": "Reference package must have no imports from src/, research/, or Stage-1 MHRN runner modules and must not contain canonical EVID/experiment identifiers or frozen target-table labels.",
+      "status": "PENDING"
+    },
+    "seed_freshness": {
+      "required": true,
+      "rule": "Evaluation seeds 810000001-810000020 must be rechecked automatically at the freeze commit; any occurrence outside the preregistration and blinded reference_protocol.json blocks freeze.",
+      "status": "PENDING"
+    },
+    "reset_parity": {
+      "required": true,
+      "status": "PENDING",
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-RESET-R1/result.json"
+    },
+    "synapse_delay_parity": {
+      "required": true,
+      "status": "PENDING",
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-SYNAPSE-R1/result.json"
+    },
+    "runner_and_verifier": {
+      "required": true,
+      "status": "IMPLEMENTED_PENDING_FREEZE_HASH_BINDING",
+      "runner": "reference/stage1_topology_brian2/runner.py",
+      "verifier": "scripts/verify_stage1_topology_reference_r1.py",
+      "rule": "Runner must pass independence/blinding scan; both files must exist before freeze and their SHA-256 digests must be written into the freeze record."
+    }
+  },
+  "reference_implementation_authorship": {
+    "project_leadership_overlap": true,
+    "classification_consequence": "Even if a separate framework/codebase is used, this project-side implementation is PARTIAL_REPLICATION only. It cannot be relabeled EXTERNAL_REPLICATION without genuinely external authorship/control and a separately registered protocol."
+  },
+  "freeze_authorization": {
+    "allowed": true,
+    "rule": "All pre-freeze gates passed and are bound into the freeze record. Reference evaluation remains unauthorized until the separate runner implementation is independently scanned, hash-bound and explicitly authorized.",
+    "reference_runner_implementation_allowed": true,
+    "reference_runner_execution_allowed": false
+  },
+  "invalidated_freeze_history": [
+    {
+      "archived_record": "research/audits/invalidated/PREREG-S1-TOPO-REFERENCE-R1.freeze.68cc945.invalidated.json",
+      "original_freeze_commit": "68cc945db5ed0822490e63a10417dfb152aa8539",
+      "status": "INVALIDATED_PREMATURE_FREEZE",
+      "reference_data_created": false,
+      "replication_credit_awarded": false
+    }
+  ],
+  "freeze_record": "research/preregistrations/PREREG-S1-TOPO-REFERENCE-R1.freeze.json",
+  "freeze_semantics": "Scientific translation, targets, equivalence bounds, endpoints and decision criteria are frozen. Reference evaluation remains forbidden until a separately reviewed runner implementation is hash-bound and explicitly authorized."
+}

research/preregistrations/PREREG-S1-TOPO-REFERENCE-R2.json

@@ -0,0 +1,557 @@
+{
+  "schema_version": 1,
+  "preregistration_id": "PREREG-S1-TOPO-REFERENCE-R2",
+  "stage": 1,
+  "research_question": "RQ-SNN-003",
+  "hypothesis": "H-SNN-003-B",
+  "claim_id": "CLAIM-S1-TOPO-001",
+  "protocol": "stage1_topology_reference_replication_r2",
+  "mode": "PROSPECTIVE_CROSS_IMPLEMENTATION_PARTIAL_REPLICATION",
+  "status": "DRAFT_PRE_FREEZE_GATES_PENDING",
+  "execution_authorized": false,
+  "scientific_evidence": false,
+  "automatic_evidence_promotion": false,
+  "replication_credit_target": "partial_only",
+  "maturity_target_if_successful": {
+    "independent_replication_component": "partial",
+    "contribution": 0.075,
+    "stage1_maturity": "85% -> 92.5%",
+    "full_external_replication_still_required_for_100_percent": true
+  },
+  "independence_contract": {
+    "framework": "Brian2",
+    "framework_version": "2.10.1",
+    "python_version": "3.13",
+    "implementation_location": "reference/stage1_topology_brian2/",
+    "separate_package_required": true,
+    "forbidden_import_prefixes": [
+      "src",
+      "scripts.run_stage1_topology",
+      "research"
+    ],
+    "forbidden_runtime_dependencies": [
+      "MHRN NeuralNetwork",
+      "MHRN Neuron",
+      "MHRN Synapse",
+      "MHRN topology builders",
+      "MHRN experiment runner modules"
+    ],
+    "allowed_shared_inputs": [
+      "frozen numeric neuron/synapse/stimulus translation contract from this preregistration",
+      "frozen topology shapes and graph rules described textually in this preregistration",
+      "evaluation seed list"
+    ],
+    "runner_blinding": "The reference runner must not contain or load the canonical MHRN effect sizes, equivalence bounds, EVID files or historical experiment statistics. Replication classification is performed only by a separate verifier after raw reference DATA are persisted.",
+    "authorship_limit": "The implementation remains under the same project leadership and therefore can earn only partial cross-implementation replication credit, never full independent external replication.",
+    "implementer": "same MHRN project leadership unless an external implementation is explicitly documented later",
+    "credit_semantics": "Because the current implementation is project-side, success can only be PARTIAL_REPLICATION. A later genuinely external implementation must use a separately registered protocol and cannot inherit full-credit status from this run.",
+    "automated_leakage_guard": "A pre-freeze AST/text scan must reject imports from MHRN runtime modules and references to EVID-2026-19, EXP-S1-TOPO-PROMO-R1-20260927, canonical_targets, frozen_bounds, equivalence bounds, research/experiments, or research/registry/evidence from within reference/stage1_topology_brian2/.",
+    "runner_source": "reference/stage1_topology_brian2/runner_r2.py",
+    "verifier_source": "scripts/verify_stage1_topology_reference_r2.py",
+    "execution_guard": "The Brian2 R2 runner refuses execution unless MHRN_STAGE1_REFERENCE_EXECUTION=AUTHORIZED_REFERENCE_R2 is explicitly supplied by a separately approved execution workflow."
+  },
+  "source_target": {
+    "canonical_experiment": "EXP-S1-TOPO-PROMO-R1-20260927",
+    "canonical_evidence": "EVID-2026-19",
+    "canonical_operating_envelope": {
+      "neurons": 64,
+      "edges": 246,
+      "dt_ms": 1,
+      "delay_ticks": 1,
+      "synaptic_weight": 55,
+      "stimulus_current": 100,
+      "stimulus_ticks": 1,
+      "evaluation_ticks": 128,
+      "auc_window_ticks": 32,
+      "input_labels": [
+        0,
+        1,
+        2,
+        3
+      ],
+      "output_labels": [
+        60,
+        61,
+        62,
+        63
+      ]
+    },
+    "conditions": {
+      "1d": [
+        64,
+        1,
+        1,
+        1,
+        1
+      ],
+      "2d": [
+        8,
+        8,
+        1,
+        1,
+        1
+      ],
+      "3d": [
+        4,
+        4,
+        4,
+        1,
+        1
+      ],
+      "5d": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "5d_shuffled": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ],
+      "random_graph": [
+        4,
+        2,
+        2,
+        2,
+        2
+      ]
+    }
+  },
+  "translation_contract": {
+    "neuron_model": "Izhikevich 2003 regular-spiking state variables v,u",
+    "initial_state": {
+      "v": -65,
+      "u": -13
+    },
+    "parameters": {
+      "a": 0.02,
+      "b": 0.2,
+      "c": -65,
+      "d": 8,
+      "threshold_mV": 30
+    },
+    "integration": "At each 1 ms tick, use the MHRN canonical two-half-Euler update for v followed by one Euler update for u: v += 0.5*(0.04*v^2+5*v+140-u+I), repeated twice; u += a*(b*v-u). A spike is detected after integration when v>=30, then reset v=c and u=u+d.",
+    "adaptation_and_secondary_mechanisms": {
+      "threshold_adaptation": {
+        "status": "ACTIVE_AND_SPIKE_TIMING_RELEVANT",
+        "parameters": {
+          "rate": 0.01,
+          "decay": 0.999,
+          "base_izhikevich_threshold": 30
+        },
+        "rule": "Spike detection uses 30.0 + threshold_adaptation. On spike, threshold_adaptation += 0.01; after firing-rate update it is multiplied by 0.999; homeostasis is then applied.",
+        "reference_requirement": "must be translated exactly"
+      },
+      "homeostasis": {
+        "status": "ACTIVE_AND_SPIKE_TIMING_RELEVANT",
+        "parameters": {
+          "target_rate_hz": 10,
+          "firing_rate_tau_ms": 1000,
+          "learning_rate": 0.001
+        },
+        "rule": "After each tick, firing_rate_estimate is updated with alpha=1-exp(-elapsed_ms/1000); homeostasis then adds 0.001*(firing_rate_estimate-10) to threshold_adaptation and clamps it to [-10,10].",
+        "reference_requirement": "must be translated exactly"
+      },
+      "energy": {
+        "status": "ACTIVE_STATE_BUT_SPIKE_TIMING_INERT_IN_CANONICAL_CORE",
+        "parameters": {
+          "initial": 1,
+          "spike_cost": 0.001,
+          "recovery_rate": 0.0001
+        },
+        "rule": "Energy is decremented on spikes and recovers after each tick, but Neuron.step/integrate_membrane/spike_threshold do not read energy to determine v/u or spike threshold.",
+        "reference_requirement": "may be omitted from the reference dynamics only if the pre-freeze source audit remains green; omission must be declared in provenance"
+      },
+      "traces": {
+        "status": "ACTIVE_STATE_BUT_SPIKE_TIMING_INERT_IN_CANONICAL_CORE",
+        "parameters": {
+          "decay": 0.95,
+          "increment": 1
+        },
+        "rule": "Pre/post traces increment/decay but are not read by membrane integration or spike threshold in the canonical topology runner.",
+        "reference_requirement": "may be omitted from the reference dynamics only if the pre-freeze source audit remains green"
+      },
+      "synaptic_plasticity": {
+        "status": "NOT_APPLIED_BY_CANONICAL_NETWORK_STEP",
+        "rule": "The topology runner creates fixed-weight Synapse objects. NeuralNetwork.step queues and delivers connection.weight but does not invoke pair-STDP, eligibility or reward updates.",
+        "reference_requirement": "reference synaptic weights remain fixed during each run"
+      },
+      "refractory": {
+        "status": "INACTIVE",
+        "refractory_ticks": 0,
+        "reference_requirement": "no additional refractory period"
+      }
+    },
+    "synapses": "Current-based excitatory events of fixed weight 55.0 with exactly one-tick delay. No plasticity during the run.",
+    "tick_order": "For tick t: clear/read external currents; deliver events whose delivery_tick==t; combine external+synaptic current; execute two half-Euler v updates and one u Euler update; test v >= 30 + current threshold_adaptation; if spiked reset v=c,u+=d and add threshold_adaptation_rate; update firing-rate estimate; decay threshold adaptation; apply homeostasis; queue outgoing fixed-weight events for t+delay. Energy/traces may update as audited but are spike-timing inert.",
+    "stimulus": "At tick 0 only, inject current 100.0 into logical labels 0,1,2,3.",
+    "topology_generation": {
+      "coordinate_order": "Generate all Cartesian-product coordinates for the declared 5D shape, then sort exactly by (sum(value_i/max(size_i-1,1)), coordinate_tuple). Logical neuron labels are assigned only after this sort. This executable ordering is the canonical R2 mapping rule.",
+      "future_only_edges": true,
+      "out_degree_cap": 4,
+      "edge_budget": 246,
+      "geometric_rule": "For each source label, connect to the up to four future labels with smallest normalized Euclidean coordinate distance; ties are resolved by a frozen seed-derived deterministic tie key, then label.",
+      "5d_shuffled": "Shuffle coordinate-to-label assignment using the frozen reference implementation's independently specified deterministic PRNG mapping before applying the same geometric edge rule.",
+      "random_graph": "For each source, choose the same number of future targets uniformly without replacement using the frozen reference PRNG mapping.",
+      "self_connections": false,
+      "parallel_connections": false,
+      "coordinate_order_correction": {
+        "date": "2026-09-27",
+        "status": "R2_SUPERSEDING_CORRECTION_BEFORE_REFERENCE_DATA",
+        "r1_issue": "R1 freeze was aborted before DATA because the frozen contract provenance did not make the executable coordinate-to-label mapping sufficiently unambiguous.",
+        "canonical_source": "scripts/run_stage1_topology_v2.py::canonical_coords",
+        "executable_gate": "research/calibrations/CAL-S1-TOPO-REFERENCE-TOPOLOGY-MAP-R2/result.json",
+        "data_seen": false
+      }
+    },
+    "metric_translation": {
+      "active_neuron": "a neuron that emitted at least one spike by the end of the evaluated tick",
+      "final_active_fraction": "number of ever-active neurons divided by 64",
+      "activation_auc_0_32": "sum of cumulative ever-active-neuron fraction over ticks 0..31 inclusive",
+      "first_output_latency_censored": "first tick with any spike from labels 60..63; if absent use the frozen censor sentinel 129",
+      "delivered_events": "count of synaptic spike events delivered to postsynaptic neurons",
+      "total_spikes": "total emitted spikes across all neurons"
+    },
+    "deviation_policy": "Any discovered semantic mismatch that could change spike timing, event delivery or registered endpoints blocks execution and requires a new preregistration revision before reference DATA are generated."
+  },
+  "evaluation": {
+    "seeds": [
+      820000001,
+      820000002,
+      820000003,
+      820000004,
+      820000005,
+      820000006,
+      820000007,
+      820000008,
+      820000009,
+      820000010,
+      820000011,
+      820000012,
+      820000013,
+      820000014,
+      820000015,
+      820000016,
+      820000017,
+      820000018,
+      820000019,
+      820000020
+    ],
+    "seed_selection_rationale": "The originally drafted block 1200021-1200040 failed the executable freshness gate because seed 1200035 already existed in historical Batch artifacts. Before any reference DATA and before freeze, the entire block was discarded. The replacement block 810000001-810000020 was selected as a new contiguous bookkeeping block and returned no repository-search matches at selection time; the executable freshness gate must still pass at the exact freeze commit.",
+    "conditions": [
+      "1d",
+      "2d",
+      "3d",
+      "5d",
+      "5d_shuffled",
+      "random_graph"
+    ],
+    "total_runs": 120,
+    "primary_endpoint": "first_output_latency_censored",
+    "secondary_endpoint": "activation_auc_0_32",
+    "primary_contrasts": [
+      [
+        "1d",
+        "2d"
+      ],
+      [
+        "2d",
+        "3d"
+      ],
+      [
+        "3d",
+        "5d"
+      ],
+      [
+        "5d",
+        "5d_shuffled"
+      ],
+      [
+        "5d",
+        "random_graph"
+      ]
+    ],
+    "primary_difference": "right-minus-left",
+    "primary_test": "exact two-sided paired sign test within reference DATA",
+    "primary_multiple_testing": "one Holm family across the five first-output-latency contrasts",
+    "alpha": 0.05,
+    "effect_interval": "deterministic paired bootstrap 95% CI for median difference",
+    "secondary_role": "activation_auc_0_32 direction is a secondary concordance check and cannot rescue failed primary replication criteria",
+    "seed_revision_history": [
+      {
+        "date": "2026-09-27",
+        "old_block": "1200021-1200040",
+        "reason": "Executable pre-freeze freshness gate found historical collision for 1200035 in EXP-BATCH-20260914074039-77 artifacts.",
+        "action": "Discarded the entire block before freeze and before any reference evaluation; no outcomes existed or were inspected.",
+        "new_block": "810000001-810000020"
+      }
+    ]
+  },
+  "canonical_targets": {
+    "source": "EXP-S1-TOPO-PROMO-R1-20260927 analysis/statistics.json",
+    "first_output_latency_median_deltas": {
+      "1d_to_2d": -10,
+      "2d_to_3d": -3,
+      "3d_to_5d": -1,
+      "5d_to_5d_shuffled": -3,
+      "5d_to_random_graph": -2.5
+    },
+    "equivalence_bounds_rule": "Strict sign-preserving equivalence fixed at 50%-150% of the absolute canonical median effect for every primary latency contrast. No resolution-based widening is used.",
+    "frozen_bounds": {
+      "1d_to_2d": [
+        -15,
+        -5
+      ],
+      "2d_to_3d": [
+        -4.5,
+        -1.5
+      ],
+      "3d_to_5d": [
+        -1.5,
+        -0.5
+      ],
+      "5d_to_5d_shuffled": [
+        -4.5,
+        -1.5
+      ],
+      "5d_to_random_graph": [
+        -3.75,
+        -1.25
+      ]
+    },
+    "activation_auc_direction_only": {
+      "1d_to_2d": "positive",
+      "2d_to_3d": "positive",
+      "3d_to_5d": "negative",
+      "5d_to_5d_shuffled": "negative",
+      "5d_to_random_graph": "negative"
+    },
+    "direction_rule": "Each primary contrast must separately retain the canonical non-zero sign and its median must lie inside the frozen strict 50%-150% equivalence interval. For 3d->5d this means median <0 and within [-1.5,-0.5].",
+    "resolution_note": "The primary latency metric is discrete and paired medians with n=20 can occur in 0.5-tick increments. This discreteness is treated as a known sensitivity limitation, not as grounds to widen the equivalence interval. A Brian2 median of -2.0 for 3d->5d is outside the preregistered equivalence bound.",
+    "other_bounds_unchanged": true,
+    "criterion_revision_history": [
+      {
+        "date": "2026-09-27",
+        "status": "DRAFT_ONLY_WITHDRAWN_BEFORE_REFERENCE_DATA",
+        "change": "A temporary DRAFT considered widening 3d->5d from [-1.5,-0.5] to [-2.0,-0.5] for metric discreteness.",
+        "resolution": "Withdrawn after methodological review before any Brian2 reference evaluation DATA existed. Canonical freeze candidate restored the original strict 50%-150% interval [-1.5,-0.5].",
+        "data_seen": false
+      }
+    ]
+  },
+  "decision_rule": {
+    "partial_replication": [
+      "reference design integrity passes",
+      "all five first-output-latency reference contrasts retain the canonical direction",
+      "all five reference median deltas lie inside their frozen equivalence bounds",
+      "all five primary paired sign tests remain Holm-significant with bootstrap CI excluding zero",
+      "at least four of five activation_auc_0_32 secondary contrast directions agree with the canonical direction"
+    ],
+    "failed_replication": [
+      "any primary latency contrast reverses canonical direction",
+      "or two or more primary latency contrasts have median effects outside the frozen equivalence bounds and their 95% bootstrap CIs are wholly outside those bounds",
+      "or fewer than three of five primary latency contrasts are Holm-significant"
+    ],
+    "inconclusive": [
+      "a semantic translation mismatch is discovered after freeze",
+      "or one or more primary latency CIs span an equivalence boundary such that success/failure classification is unstable",
+      "or exactly three or four primary contrasts satisfy both Holm significance and equivalence while none reverse direction",
+      "or exactly one primary contrast is wholly outside its equivalence bound without direction reversal while the remaining four satisfy the primary criteria",
+      "or reference runtime behavior makes a preregistered endpoint non-comparable"
+    ],
+    "equal_value_rule": "PARTIAL_REPLICATION, FAILED_REPLICATION and INCONCLUSIVE_REFERENCE_REPLICATION are equally reportable scientific outcomes. No secondary endpoint or post-hoc tuning may rescue a failed primary classification."
+  },
+  "provenance_requirements": {
+    "separate_code_digest": true,
+    "framework_version_recorded": true,
+    "python_version_recorded": true,
+    "dependency_lock_recorded": true,
+    "clean_git_tree": true,
+    "runtime_and_platform_recorded": true,
+    "raw_reference_data_persisted": true,
+    "runner_source_hash_persisted": true,
+    "verifier_source_hash_persisted": true,
+    "no_shared_mhrn_code_imports_verified": true
+  },
+  "maturity_semantics": {
+    "success": "A successful reference replication counts only as partial independent/reference replication = 7.5/15 points, advancing Stage-1 maturity from 85% to 92.5%.",
+    "failure": "A failed or inconclusive reference replication does not increase maturity and must remain visible in the canonical Stage-1 state.",
+    "full_credit": "15/15 replication credit remains reserved for genuinely external, independently controlled/authored replication."
+  },
+  "review": {
+    "human_review_required": true,
+    "automatic_evidence_promotion": false,
+    "execution_creates_reference_data_only": true
+  },
+  "prefreeze_requirements": {
+    "integrator_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-INTEGRATOR-R1/result.json",
+      "protocol": "one neuron, one tick, current=0 and current=100; compare MHRN and Brian2 final v,u, threshold_adaptation, firing_rate_estimate, spike flag",
+      "tolerance_abs": 1e-12,
+      "pass_rule": "all compared scalar fields match within 1e-12 and spike flags are identical in both cases",
+      "failure_rule": "freeze blocked; no reference evaluation may run"
+    },
+    "mechanism_audit": {
+      "required": true,
+      "artifact": "research/audits/STAGE1_TOPOLOGY_REFERENCE_MECHANISM_AUDIT_20260927.json",
+      "pass_rule": "every spike-timing-relevant canonical mechanism is classified and translated; every omitted mechanism is source-demonstrated spike-timing inert",
+      "failure_rule": "freeze blocked"
+    },
+    "runner_independence_scan": {
+      "required": true,
+      "script": "scripts/check_stage1_reference_independence.py",
+      "pass_rule": "no forbidden MHRN imports, EVID/experiment access, canonical target/equivalence access, or research artifact reads in the reference runner package",
+      "failure_rule": "freeze blocked"
+    },
+    "seed_freshness": {
+      "required": true,
+      "script": "scripts/check_stage1_reference_seed_freshness_r2.py",
+      "pass_rule": "R2 evaluation seeds must be absent from prior DATA/experiment use and may occur only in the R2 preregistration, sanitized R2 protocol, and after freeze the R2 freeze record.",
+      "failure_rule": "select a new seed block before freeze",
+      "allowed_seed_declarations": [
+        "research/preregistrations/PREREG-S1-TOPO-REFERENCE-R2.json",
+        "reference/stage1_topology_brian2/reference_protocol_r2.json"
+      ]
+    },
+    "reset_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-RESET-R1/result.json",
+      "protocol": "one neuron, forced spike followed by the next tick; compare spike flag, reset v=c, u increment by d, threshold adaptation, firing-rate estimate and next-tick state",
+      "tolerance_abs": 1e-12,
+      "pass_rule": "all compared state fields and spike decisions match MHRN within 1e-12",
+      "failure_rule": "freeze blocked"
+    },
+    "synapse_delay_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-SYNAPSE-R1/result.json",
+      "protocol": "two neurons, one fixed excitatory synapse weight 55 and delay 1 tick; induce a source spike at tick 0 and compare event delivery tick, delivered-event count, postsynaptic current/state and spike decision",
+      "tolerance_abs": 1e-12,
+      "pass_rule": "event is absent at source-spike tick, delivered exactly one tick later, and compared target state/spike fields match MHRN within 1e-12",
+      "failure_rule": "freeze blocked"
+    },
+    "topology_mapping_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-TOPOLOGY-MAP-R2/result.json",
+      "protocol": "Across every registered condition and three non-evaluation probe seeds, compare independently generated coordinate lists and complete directed edge lists between canonical MHRN topology semantics and the Brian2 R2 reference implementation.",
+      "pass_rule": "All coordinate lists and complete edge lists match exactly for every condition/probe-seed pair; source hashes are recorded.",
+      "failure_rule": "R2 freeze and all reference evaluation are blocked."
+    },
+    "dynamics_reuse_validation": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-DYNAMICS-REUSE-R2/result.json",
+      "script": "scripts/verify_stage1_reference_dynamics_reuse_r2.py",
+      "pass_rule": "Historical R1 integrator/reset/synapse artifacts remain byte-preserved; all recorded source hashes still match current dynamics sources and all historical parity results remain PASS under Brian2 2.10.1.",
+      "failure_rule": "R2 freeze blocked; create new R2-specific dynamics calibrations rather than modifying R1 history."
+    }
+  },
+  "prefreeze_review_record": {
+    "status": "METHOD_CORRECTION_REGISTERED_PRE_DATA",
+    "review_basis": "research/decisions/2026-09-27_stage1_reference_r1_prefreeze_deviation.json",
+    "human_execution_authorization": "NOT_GRANTED",
+    "reference_data_seen": false
+  },
+  "pre_freeze_gates": {
+    "integrator_parity": {
+      "required": true,
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-INTEGRATOR-R1/result.json",
+      "rule": "PASS required before freeze. Single-tick subthreshold/spike cases and persistent multi-tick trajectory must match v/u/adaptation/firing-rate state at absolute tolerance 1e-12; the multi-tick trajectory must include at least one spike.",
+      "status": "REUSE_REQUIRES_SOURCE_HASH_REVALIDATION"
+    },
+    "mechanism_audit": {
+      "required": true,
+      "artifact": "research/audits/2026-09-27_stage1_reference_mechanism_audit.md",
+      "rule": "Threshold adaptation and homeostasis are timing-relevant and must be translated. Energy and traces may be omitted only as explicitly audited causally inert state for registered propagation endpoints.",
+      "status": "PASS_DOCUMENTED"
+    },
+    "independence_scan": {
+      "required": true,
+      "script": "scripts/check_stage1_reference_independence.py",
+      "rule": "Reference package must have no imports from src/, research/, or Stage-1 MHRN runner modules and must not contain canonical EVID/experiment identifiers or frozen target-table labels.",
+      "status": "PENDING_R2"
+    },
+    "seed_freshness": {
+      "required": true,
+      "rule": "Evaluation seeds 810000001-810000020 must be rechecked automatically at the freeze commit; any occurrence outside the preregistration and blinded reference_protocol.json blocks freeze.",
+      "status": "PENDING_R2"
+    },
+    "reset_parity": {
+      "required": true,
+      "status": "REUSE_REQUIRES_SOURCE_HASH_REVALIDATION",
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-RESET-R1/result.json"
+    },
+    "synapse_delay_parity": {
+      "required": true,
+      "status": "REUSE_REQUIRES_SOURCE_HASH_REVALIDATION",
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-SYNAPSE-R1/result.json"
+    },
+    "runner_and_verifier": {
+      "required": true,
+      "status": "IMPLEMENTED_PENDING_R2_FREEZE_HASH_BINDING",
+      "runner": "reference/stage1_topology_brian2/runner_r2.py",
+      "verifier": "scripts/verify_stage1_topology_reference_r2.py",
+      "rule": "Runner must pass independence/blinding scan; runner, verifier and contract tests must be hash-bound into the R2 freeze record."
+    },
+    "topology_mapping_parity": {
+      "required": true,
+      "status": "PENDING_R2",
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-TOPOLOGY-MAP-R2/result.json"
+    },
+    "dynamics_reuse_validation": {
+      "required": true,
+      "status": "PENDING_R2",
+      "artifact": "research/calibrations/CAL-S1-TOPO-REFERENCE-DYNAMICS-REUSE-R2/result.json"
+    }
+  },
+  "reference_implementation_authorship": {
+    "project_leadership_overlap": true,
+    "classification_consequence": "Even if a separate framework/codebase is used, this project-side implementation is PARTIAL_REPLICATION only. It cannot be relabeled EXTERNAL_REPLICATION without genuinely external authorship/control and a separately registered protocol."
+  },
+  "freeze_authorization": {
+    "allowed": false,
+    "rule": "R2 may freeze only after all source-bound dynamics gates, R2 independence/seed freshness, and exact topology-mapping parity pass.",
+    "reference_runner_implementation_allowed": true,
+    "reference_runner_execution_allowed": false
+  },
+  "invalidated_freeze_history": [
+    {
+      "archived_record": "research/audits/invalidated/PREREG-S1-TOPO-REFERENCE-R1.freeze.68cc945.invalidated.json",
+      "original_freeze_commit": "68cc945db5ed0822490e63a10417dfb152aa8539",
+      "status": "INVALIDATED_PREMATURE_FREEZE",
+      "reference_data_created": false,
+      "replication_credit_awarded": false
+    }
+  ],
+  "freeze_record": "research/preregistrations/PREREG-S1-TOPO-REFERENCE-R2.freeze.json",
+  "freeze_semantics": "R2 is not frozen yet. Its corrected topology mapping and all comparison criteria must be frozen before any reference evaluation DATA. R1 remains historical and may never authorize execution.",
+  "supersedes": "PREREG-S1-TOPO-REFERENCE-R1",
+  "supersession": {
+    "date": "2026-09-27",
+    "reason": "R1 was aborted before reference DATA after a code-backed audit found that the frozen topology coordinate-order wording was semantically ambiguous relative to the canonical MHRN label-order implementation.",
+    "r1_freeze_record": "research/preregistrations/PREREG-S1-TOPO-REFERENCE-R1.freeze.json",
+    "r1_deviation_record": "research/decisions/2026-09-27_stage1_reference_r1_prefreeze_deviation.json",
+    "reference_data_created_under_r1": false,
+    "replication_credit_awarded_under_r1": false,
+    "scientific_evidence_under_r1": false,
+    "policy": "R1 remains immutable historical provenance. R2 is the only protocol eligible for any future reference execution."
+  },
+  "execution_authorization_requirements": {
+    "current_value": false,
+    "automatic_authorization_forbidden": true,
+    "all_required": [
+      "canonical R2 freeze record exists and all bound hashes match",
+      "R2 runner independence/blinding scan passes at the exact freeze commit",
+      "R2 seed-freshness gate passes at the exact freeze commit",
+      "R2 immutable dynamics-reuse validation passes with source hashes matching the historical R1 parity artifacts",
+      "R2 topology-mapping parity passes for all registered conditions",
+      "no R1 or R2 reference evaluation DATA exist before authorization",
+      "verifier decision-rule contract tests match the frozen R2 preregistration",
+      "an explicit separate human execution-authorization record names the frozen commit and runner hash"
+    ],
+    "human_authority": "Thomas Heisig or a separately documented authorized human reviewer"
+  }
+}