Layers 1–4

Claim Ledger

Layers 1-4: claim-level uncertainty quantification and deterministic gate enforcement.

Claim Ledger decomposes every response into atomic claims, verifies each claim against authoritative sources, scores uncertainty across up to four layers, and enforces a deterministic emission gate.

Without this

Without claim-level decomposition, you’re scoring responses as monoliths. One backed claim hides three fabricated ones. You can’t tell auditors which specific assertion failed or why.

Position in the platform

System	Layer	Function
Oracle Foundry	Foundation	Supplies authoritative chunks with provenance, tiering, and SIRE boundaries
SIRE Crosswalk	Post-Foundry	Provides global authority routing and conflict context for cross-oracle claims
Prompt Compiler	L0	Optimizes prompt structure before generation
Claim Ledger	L1-L4	Scores per-claim confidence, entailment, stability, and representation uncertainty
Process Control System	Cross-layer	Tracks drift and process capability over governed output
Forensics Lab	L5	Runs root-cause diagnostics on failed governed responses

Five-stage runtime pipeline

Stage 1

Claim extraction

Splits model output into atomic factual assertions with claim IDs and claim types.

Stage 2

L1 calibrated confidence

Converts token logprobs into calibrated correctness probabilities, with separate scoring for critical tokens.

Stage 3

L2 source entailment

Scores each claim against oracle chunks using embedding similarity + NLI, then applies tier-weighted aggregation.

Stage 4

L3 stability

Uses multi-draw regeneration and semantic clustering to identify reproducible claims vs stochastic artifacts.

Stage 5

L4 representation uncertainty

Applies hidden-state probes to detect internal volatility not visible in token confidence.

Deterministic gate logic

Authority boundary check (SIRE exclusions and out-of-scope concepts)

License boundary check for output constraints on licensed material

Worst-claim entailment gate using min(best_score) across all claims

Contradiction detection (numeric predicates, polarity, and NLI where required)

Mode-dependent confidence, stability, and representation uncertainty checks

Governance envelope output

• Per-claim audit trail with source attribution, layer scores, and gate reason codes
• Complete request provenance: model, oracle, embedding, calibration, and compiler versions
• Structured artifact consumed by Process Control System and Forensics Lab for closed-loop remediation

Availability by deployment model

API without logprobs

L2 only

Entailment-based governance with claim extraction, source scoring, and gate outcomes.

API with logprobs

L1 + L2

Adds calibrated confidence on critical tokens to source entailment enforcement.

Self-hosted standard

L1 + L2 + L3

Adds stability analysis using controlled multi-draw generation.

Self-hosted full

L1 + L2 + L3 + L4

Activates full four-layer claim governance with hidden-state uncertainty probes.

Quality economics

Claim Ledger computes per-claim and per-request Taguchi loss so governance quality is measured as an economic signal, not just a pass/fail label.

Live emission gate result

Actual EmissionGateResult from a governed HIPAA response with five extracted claims.

{  "outcome": "warn_only",  "mode": "advisory",  "blocked": false,  "blocked_reason": null,  "candidate_claim_count": 5,  "failed_claim_count": 0,  "warning_claim_count": 1,  "failed_rules": [],  "warning_rules": ["missing_citation"],  "warning_claim_ids": ["clm_04"],  "claim_verdicts": [    {      "claim_id": "clm_01",      "verdict": "supported",      "citations": ["E3", "E7"]    },    {      "claim_id": "clm_02",      "verdict": "supported",      "citations": ["E3"]    },    {      "claim_id": "clm_03",      "verdict": "partial",      "citations": ["W2"]    },    {      "claim_id": "clm_04",      "verdict": "uncited",      "citations": []    },    {      "claim_id": "clm_05",      "verdict": "supported",      "citations": ["E7", "E12"]    }  ]}

How we look inside the model

Claim Ledger’s per-claim scores come from real inference-time signals: token logprobs, entailment scoring, stability analysis, and hidden-state probes. We call it AI Neurosurgery — because the black box was never actually closed.

See how it works →

Govern claims, not responses

Response-level scoring hides failure mechanisms. Claim Ledger isolates each assertion, enforces deterministic thresholds, and records a reproducible evidence chain.

Architecture Prompt Compiler SIRE Crosswalk Process Control System Forensics Lab The Clean Room Check Your Risk

Who uses this

Operator

Governance runtime

Executes automatically on every governed request—decomposes, scores, and gates without manual review.

Consumer

Compliance officers and auditors

Consume per-claim audit trails. Process Control System and Forensics Lab use gate outcomes for statistical monitoring and root-cause analysis.