AXIOM-XDOMAIN-002 — Preregistered Replication Experiment (PRIMARY)

[BordneAI]

AXIOM-XDOMAIN-002: Preregistered Replication Experiment (PRIMARY)

Purpose

This is the canonical tracking issue for AXIOM-XDOMAIN-002.

AXIOM-XDOMAIN-002 is a preregistered replication experiment designed to test whether independently logged predictions and controlled computational baselines show any meaningful cross-domain signal when compared against independently ingested physical/publication events.

This issue tracks the full experiment lifecycle across the preserved 7-phase model:

1. Foundation / Lock
2. Baseline Data Collection
3. Dataset Integrity & Separation
4. Event Ingestion
5. Outcome Adjudication
6. Scoring / Reporting
7. Evaluation / Promotion Review

This issue must remain open until data collection, event ingestion, adjudication, scoring/reporting, and final evaluation are complete.

---

Current Project Status

Status: Blocked

Reason:

• Experiment execution is active.
• Completion and evaluation are blocked by missing live physical/publication event ingestion, physical matching, final adjudication, final scoring/reporting, and Phase 7 evaluation.
• No T2_candidate review is actionable until final evaluation exists.

This issue is blocked at the lifecycle-completion level, even though some execution phases are active or in review.

---

Current Verified System State

Implemented or active:

• Experiment scaffold exists
• Preregistration and lock materials are established
• Scoring lock exists
• Monitoring/audit log stream exists
• Cognitive prediction logging sub-issue completed
• Scheduled computational baseline sub-issue completed
• Monitoring/audit log initialization sub-issue completed
• Production computational cadence is active
• LM Studio cadence and JSON handling have been stabilized
• Computational outputs are being generated
• Computational normalization is implemented
• Baseline scoring output exists
• Fallback/generated recovery outputs are excluded from scoring
• Valid computational baseline outputs are retained as null baseline records
• Negative-null / exclusion handling is partially implemented
• Physical ingestion implementation is in progress and ready for first real independent source
• Fixture-safe physical event append/adjudication path has been verified without mutating the live event ledger
• Non-scoring adjudication bridge is implemented and in review
• Physical-event adjudication fixture test is implemented
• Cadence health-check script exists and has reported pass status
• Interim report generator is implemented and in review
• Interim report explicitly preserves baseline-only interpretation

Not yet complete:

• No real physical/publication event has been appended to the live event ledger
• Physical matching has not been performed
• Final window-aware adjudication is not complete
• Final topic-aware adjudication is not complete
• Final null/miss/ambiguous outcome closure is not complete
• Final scoring run is not complete
• Final report is not complete
• Phase 7 evaluation is not complete
• T2_candidate assessment is blocked

---

Latest Baseline Snapshot

Current latest reported snapshot:

• health-check status: pass
• event ledger total: 15
• cognitive events: 5
• computational events: 10
• physical events: 0
• scoring records: 10
• scored records: 8
• scoring exclusions: 2
• adjudication prediction records: 15
• adjudication exclusions: 3
• physical matching performed: false
• scoring performed in adjudication bridge: false
• signal conclusion: none
• T2_candidate assessment: blocked

Interpretation:

• Current results are baseline-only.
• No qualifying live physical/publication event has been ingested.
• No physical matching has been performed.
• No hit/miss/signal conclusion can be drawn yet.
• No positive or negative anomaly finding should be inferred at this stage.
• T2_candidate review is not actionable until final evaluation exists.

---

Dependencies

Required dependencies:

• Add and stabilize automated monitoring workflow (cron + LM Studio cadence) #44: Add and stabilize automated monitoring workflow (cron + LM Studio cadence)
• Add AXIOM-XDOMAIN README and usage docs #46: Add AXIOM-XDOMAIN README and usage docs
• Add experimental dataset labeling + isolation #45: Add experimental dataset labeling + isolation
• Add Physical / Cross-Domain Event Ingestion #41: Add Physical / Cross-Domain Event Ingestion
• Add Deterministic Outcome Adjudication and Scoring Pipeline #52: Add Deterministic Outcome Adjudication and Scoring Pipeline
• Add deterministic outcome adjudication and scoring pipeline #62: Add deterministic outcome adjudication and scoring pipeline

Sub-issue dependencies:

• Log preregistered predictions (cognitive) #57: Log preregistered predictions (cognitive)
• Establish scheduled computational baseline #58: Establish scheduled computational baseline
• Initialize monitoring and audit logs #59: Initialize monitoring and audit logs
• Add physical event ingestion #60: Add physical event ingestion
• Prepare outcome adjudication pipeline #61: Prepare outcome adjudication pipeline

Related downstream review:

• Define upgrade path to T2_candidate validation #47: Define upgrade path to T2_candidate validation
• Add experiment summary and export report generator #56: Add experiment summary and export report generator

---

Execution Phases

Phase 1: Foundation / Lock

Status: Complete

Completed:

• Experiment scaffold created
• Preregistration files created
• LOCK.json created
• scoring_model.lock.json created
• Lock boundary established
• Experiment activation boundary recorded
• Raw record preservation policy established
• Append-only audit expectations defined
• No reuse of XDOMAIN-001 predictions
• Experiment output remains isolated from KB promotion

Relevant issues:

• Add AXIOM-XDOMAIN experiment pipeline scaffold #38
• Add live prediction hashing logger #39
• Define AXIOM-XDOMAIN scoring model #42
• Add experiment integrity & anti-bias rules #43
• AXIOM-XDOMAIN-002 — Preregistered Replication Experiment (PRIMARY) #48
• Log preregistered predictions (cognitive) #57
• Establish scheduled computational baseline #58
• Initialize monitoring and audit logs #59

---

Phase 2: Baseline Data Collection

Status: Active

Completed or active:

• Cognitive prediction logging completed
• Scheduled computational baseline established
• Monitoring/audit log initialized
• Computational cadence running through scheduled LM Studio workflow
• Computational raw records generated
• Computational records normalized
• Initial baseline scoring output generated
• Cadence health-check script implemented
• Latest health-check status reported as pass

Still required:

• Continue scheduled computational cadence
• Preserve all valid, null, fallback, excluded, and malformed outputs
• Continue monitoring for missed runs, parse failures, and cadence drift
• Do not treat baseline-only results as signal

Relevant issues:

• Add and stabilize automated monitoring workflow (cron + LM Studio cadence) #44
• Harden LM Studio cadence reliability and JSON compliance #54
• Add monitoring and alerting for cadence failures #55
• Establish scheduled computational baseline #58

---

Phase 3: Dataset Integrity & Separation

Status: Partially implemented / active

Completed or active:

• Experimental records are isolated from KB/release surfaces
• Raw computational records remain outside authoritative KB files
• Normalized records are derived experiment artifacts
• Scoring output remains an experiment artifact
• Null and exclusion handling is partially implemented
• Fallback/generated recovery outputs are excluded from scoring
• Valid no-update computational baseline records are preserved
• KB non-promotion rule is active
• README / experiment docs document KB vs experiment separation

Still required:

• Keep README / experiment docs current as implementation state changes
• Violation logging procedure must be documented
• Final reporting must include positives, nulls, exclusions, and ambiguous outcomes
• Optional validation/checking should detect accidental experiment-to-KB promotion where practical

Relevant issues:

• Add experiment integrity & anti-bias rules #43
• Add experimental dataset labeling + isolation #45
• Add AXIOM-XDOMAIN README and usage docs #46
• Enforce KB / Experiment Separation #49
• Enforce negative-null outcome handling in pipeline #53

---

Phase 4: Event Ingestion

Status: In progress

Current state:

• Physical ingestion implementation is ready for the first real independent source.
• Fixture-safe physical ingestion path has been verified.
• A fake physical event can be appended to a temporary ledger, validated, and read by the adjudication bridge without changing the live cross_domain_events.json.
• No real physical/publication event has been appended to the live ledger yet.

Still required:

• Ingest first qualifying real independent physical/publication event
• Preserve independence from prediction content
• Confirm live event record includes required fields
• Confirm source tier, confidence, topic tags, source reference, independence group, contamination risk, and negative-null handling
• Preserve append-only behavior
• Ensure event records can be consumed by adjudication/scoring pipeline

Relevant issues:

• Add physical event ingestion script #40
• Add Physical / Cross-Domain Event Ingestion #41
• Add physical event ingestion #60

Blocking effect:

• Phase 5 final adjudication remains blocked until live physical/publication event ingestion exists.
• Phase 6 final scoring/reporting remains blocked until event ingestion and matching exist.
• Phase 7 evaluation remains blocked until final adjudication and scoring exist.

---

Phase 5: Outcome Adjudication

Status: In review

Current state:

• Non-scoring adjudication bridge is implemented.
• Physical-event adjudication fixture test is implemented.
• Fixture test verifies that the bridge can ingest and represent a physical event from a temporary fixture without mutating the live experiment ledger.
• The adjudication bridge preserves raw records, applies logged exclusions, computes outcome windows, and does not perform physical matching or scoring yet.

Latest reported adjudication draft:

• prediction records: 15
• physical records: 0
• logged outcome overrides: 1
• open windows: 12
• excluded records: 3
• scoring performed: false
• physical matching performed: false

Still required:

• Physical event matching against live event records
• Window-aware adjudication
• Topic-aware adjudication
• Miss classification after outcome windows close
• Ambiguous outcome handling
• Final null result closure logic
• Reproducible adjudication output from committed files
• Confirmation that no retroactive edits occur

Relevant issues:

• Add Deterministic Outcome Adjudication and Scoring Pipeline #52
• Prepare outcome adjudication pipeline #61
• Add deterministic outcome adjudication and scoring pipeline #62

Blocking effect:

• Final scoring cannot complete until physical matching and final adjudication are implemented.
• No signal conclusion can be drawn while physical matching remains false.

---

Phase 6: Scoring / Reporting

Status: Partially implemented

Current scoring/reporting state:

• Computational normalization exists
• Baseline scoring exists
• Fallback/generated recovery outputs are excluded
• Null baseline records are preserved
• Interim report generator is implemented and in review
• Interim report is baseline-only
• Interim report does not mutate raw records
• Interim report states no physical events have been ingested, no physical matching has been performed, and no signal conclusion can be drawn

Latest reported reporting snapshot:

• health-check status: pass
• event ledger total: 15
• cognitive events: 5
• computational events: 10
• physical events: 0
• scoring records: 10
• scored records: 8
• scoring exclusions: 2
• adjudication prediction records: 15
• adjudication exclusions: 3
• physical matching performed: false
• signal conclusion: none
• T2_candidate assessment: blocked

Still required:

• Integrate physical event matching into final scoring
• Complete window-aware and topic-aware adjudication
• Finalize null baseline comparison
• Preserve hit, miss, null, ambiguous, and excluded outcomes
• Generate final scoring output
• Generate final human-readable and machine-readable reports
• Include positives, nulls, misses, exclusions, and ambiguous outcomes
• Clearly mark incomplete or blocked interpretation when applicable

Relevant issues:

• Enforce negative-null outcome handling in pipeline #53
• Add monitoring and alerting for cadence failures #55
• Add experiment summary and export report generator #56
• Add deterministic outcome adjudication and scoring pipeline #62

Blocking effect:

• Phase 7 evaluation remains blocked until final scoring and final report exist.

---

Phase 7: Evaluation / Promotion Review

Status: Blocked

Current state:

• T2_candidate path is defined but not actionable.
• T2_candidate review remains blocked until full AXIOM-XDOMAIN-002 evaluation exists.
• Current result is baseline-only.
• No qualifying physical/publication event has been adjudicated.
• No T2_candidate assessment can be made yet.

Required before Phase 7 can proceed:

• Live physical/publication event ingestion completed
• Outcome adjudication completed
• Final scoring completed
• Final report completed
• Nulls, misses, exclusions, and ambiguous outcomes included
• Statistical/comparative evaluation performed
• Contamination risk reviewed
• Source-tier quality reviewed
• Control comparison reviewed
• Governance review performed if promotion is considered

Possible final decisions:

• Remain T3 experimental screening
• Remain T3 pending more data
• T2_candidate review recommended
• Invalid / contaminated run
• New replication required

No result may be promoted directly from AXIOM-XDOMAIN output to KB fact.

Relevant issues:

• Define upgrade path to T2_candidate validation #47
• AXIOM-XDOMAIN-002 — Preregistered Replication Experiment (PRIMARY) #48
• Add experiment summary and export report generator #56

---

Constraints

The following constraints are locked for AXIOM-XDOMAIN-002:

• No reuse of XDOMAIN-001 predictions
• Predictions must be logged before outcome windows
• Computational runs must not use browsing or live external data
• Scoring model is fixed via scoring_model.lock.json
• Logs and raw records must be append-only
• Raw evidence files must not be retroactively edited
• All prediction/computational records must be hash-linked prior to outcome
• Negative-null handling must be preserved
• Fallback/generated recovery outputs must not be scored as valid predictions
• Physical/publication events must be ingested independently of prediction content
• Event matching must occur only through adjudication
• Experiment outputs must not be promoted into the KB without separate governance review

---

Canonical Artifacts

Experiment files:

• /[repo_root]/experiments/axiom_xdomain_002/LOCK.json
• /[repo_root]/experiments/axiom_xdomain_002/preregistration.json
• /[repo_root]/experiments/axiom_xdomain_002/scoring_model.lock.json
• /[repo_root]/experiments/axiom_xdomain_002/null_results.json
• /[repo_root]/experiments/axiom_xdomain_002/cross_domain_events.json
• /[repo_root]/experiments/axiom_xdomain_002/normalized_computational_events.json
• /[repo_root]/experiments/axiom_xdomain_002/scores.json
• /[repo_root]/experiments/axiom_xdomain_002/adjudication_draft.json
• /[repo_root]/experiments/axiom_xdomain_002/reports/summary.json
• /[repo_root]/experiments/axiom_xdomain_002/reports/summary.md
• /[repo_root]/experiments/axiom_xdomain_002/SOURCE_ELIGIBILITY.md

Scripts:

• /[repo_root]/scripts/axiom_xdomain_002_normalize_computational.py
• /[repo_root]/scripts/axiom_xdomain_002_score.py
• /[repo_root]/scripts/axiom_xdomain_002_add_physical.py
• /[repo_root]/scripts/axiom_xdomain_002_adjudicate.py
• /[repo_root]/scripts/axiom_xdomain_002_test_physical_adjudication_fixture.py
• /[repo_root]/scripts/axiom_xdomain_002_healthcheck.py
• /[repo_root]/scripts/axiom_xdomain_002_report.py

Raw evidence folders:

• /[repo_root]/experiments/axiom_xdomain_002/raw/cognitive/
• /[repo_root]/experiments/axiom_xdomain_002/raw/computational/

Logs:

• /[repo_root]/experiments/axiom_xdomain_002/logs/monitoring_log.jsonl

---

Current Related Issue Status

Done:

• Add AXIOM-XDOMAIN experiment pipeline scaffold #38: Add AXIOM-XDOMAIN experiment pipeline scaffold
• Add live prediction hashing logger #39: Add live prediction hashing logger
• Define AXIOM-XDOMAIN scoring model #42: Define AXIOM-XDOMAIN scoring model
• Log preregistered predictions (cognitive) #57: Log preregistered predictions (cognitive)
• Establish scheduled computational baseline #58: Establish scheduled computational baseline
• Initialize monitoring and audit logs #59: Initialize monitoring and audit logs

In review:

• Add physical event ingestion script #40: Add physical event ingestion script
• Add and stabilize automated monitoring workflow (cron + LM Studio cadence) #44: Add and stabilize automated monitoring workflow
• Add AXIOM-XDOMAIN README and usage docs #46: Add AXIOM-XDOMAIN README and usage docs
• Harden LM Studio cadence reliability and JSON compliance #54: Harden LM Studio cadence reliability and JSON compliance
• Add monitoring and alerting for cadence failures #55: Add monitoring and alerting for cadence failures
• Add experiment summary and export report generator #56: Add experiment summary and export report generator
• Prepare outcome adjudication pipeline #61: Prepare outcome adjudication pipeline

In progress:

• Add Physical / Cross-Domain Event Ingestion #41: Add Physical / Cross-Domain Event Ingestion
• Add experiment integrity & anti-bias rules #43: Add experiment integrity & anti-bias rules
• Add experimental dataset labeling + isolation #45: Add experimental dataset labeling + isolation
• Enforce KB / Experiment Separation #49: Enforce KB / Experiment Separation
• Add Deterministic Outcome Adjudication and Scoring Pipeline #52: Add Deterministic Outcome Adjudication and Scoring Pipeline
• Enforce negative-null outcome handling in pipeline #53: Enforce negative-null outcome handling in pipeline
• Add physical event ingestion #60: Add physical event ingestion
• Add deterministic outcome adjudication and scoring pipeline #62: Add deterministic outcome adjudication and scoring pipeline

Blocked:

• Define upgrade path to T2_candidate validation #47: Define upgrade path to T2_candidate validation
• AXIOM-XDOMAIN-002 — Preregistered Replication Experiment (PRIMARY) #48: AXIOM-XDOMAIN-002 — Preregistered Replication Experiment (PRIMARY)

---

Acceptance Criteria

Foundation / Lock

• Experiment scaffold created
• Preregistration files created
• Lock file committed
• Lock confirmed in monitoring log
• Experiment activation boundary recorded
• Scoring model lock present
• Raw record preservation policy established

Baseline Data Collection

• Cognitive prediction logging sub-issue completed
• Scheduled computational baseline sub-issue completed
• Monitoring/audit log sub-issue completed
• Production computational cadence verified
• Computational records normalized
• Initial baseline scoring output generated
• Fallback computational records excluded from scoring
• Valid computational baseline records retained as null baseline records
• Cadence health-check script implemented
• Continued scheduled cadence preserved through active experiment window

Dataset Integrity & Separation

• Experimental artifacts isolated from KB/release surfaces
• No automatic experiment-output to KB path allowed
• Nulls and exclusions preserved in scoring outputs
• Fallback outputs excluded from scoring
• README / experiment docs document KB vs experiment separation
• Violation logging procedure documented
• Final report includes positives, nulls, misses, exclusions, and ambiguous outcomes

Event Ingestion

• Physical ingestion implementation prepared
• Fixture-safe append/adjudication path verified
• Live event ledger preserved during fixture testing
• First real physical/publication event ingested
• Event records include required fields
• Source tier and confidence recorded
• Topic tags recorded
• Source reference recorded
• Event ingestion does not require prediction IDs
• Event ingestion does not read or modify prediction content
• Existing event records preserved
• Event records consumed by adjudication/scoring pipeline

Outcome Adjudication

• Non-scoring adjudication bridge implemented
• Fixture physical-event adjudication test implemented
• Fallback outputs classified as excluded
• Valid computational baseline records classified as null baseline records
• Excluded records preserved
• Raw computational records preserved without mutation
• Live physical event matching implemented
• Window-aware adjudication implemented
• Topic-aware adjudication implemented
• Miss classification after window closure implemented
• Ambiguous outcome handling implemented
• Final null result closure logic implemented
• Adjudication output reproducible from committed files
• No retroactive edit behavior confirmed

Scoring / Reporting

• Computational normalization implemented
• Baseline scoring output generated
• Interim report generator implemented
• Baseline-only report generated
• Interim report marks incomplete adjudication
• Interim report does not mutate raw records
• Interim report does not promote experimental outputs into KB facts
• Physical event matching integrated into final scoring
• Null baseline comparison finalized
• Final scoring run completed
• Final scoring reviewed against locked protocol
• Final report generated
• Final report includes hit/miss/null/ambiguous/excluded counts
• Final report includes T3 / T2_candidate evaluation status

Evaluation / Promotion Review

• T2_candidate threshold defined in preregistration / scoring lock
• Minimum score threshold defined
• Independent cluster requirement defined
• Control comparison requirement defined
• Pre-outcome hash requirement defined
• KB contamination prohibition defined
• AXIOM-XDOMAIN-002 dataset finalized
• Physical event ingestion completed
• Outcome adjudication completed
• Final scoring completed
• Nulls, misses, exclusions, and ambiguous outcomes included
• Statistical / comparative evaluation performed
• Final decision documented
• Governance review completed if promotion is considered

---

Notes

This is a preregistered replication experiment.

All locked rules, constraints, and scoring assumptions must be treated as immutable after activation.

Any deviation must be explicitly logged and treated as either:

• a protocol violation, or
• a trigger for a new experiment version

Current state remains baseline-only.

No live physical/publication event has been ingested yet.

No physical matching has been performed.

No hit, miss, signal, anomaly, or T2_candidate conclusion should be inferred until event ingestion, adjudication, scoring/reporting, and final evaluation are complete.

This issue should remain open and blocked until AXIOM-XDOMAIN-002 has completed all seven phases through final evaluation / promotion review.