Decoder
VERITY doesn't generate.
It compiles.

Every sentence in a VERITY response is independently measured before inclusion. Sentences that fail verification are excluded — not softened, not hedged, not qualified. The system would rather say less than say something it cannot prove.

What it is

A compilation engine,
not a generation engine.

"Language models predict the next word. VERITY predicts whether the next claim is true enough to say. The difference is everything."

Language models generate text by predicting the most statistically likely continuation of what came before. This produces fluent, confident-sounding responses regardless of whether the underlying claim is accurate. The model does not know what it knows. It knows what tends to follow what.

VERITY's decoder works differently. It does not predict words — it retrieves verified sentences and assembles them. Each candidate sentence is measured against the query before inclusion. If the fidelity score falls below the threshold for commitment, the sentence is not included. The system outputs exactly as much as the evidence supports, and no more.

What it does

Four honest answers.

The output of the decoder is not just a response — it is a response with a posture. The posture is derived from the geometric evidence, not from frequency statistics. It describes this specific output, for this specific query.

Commit
The evidence supports commitment.
All measurement dimensions agree. The response is assembled from content that passed verification at every level. Act on this output.
Caution
Most dimensions agree. One is uncertain.
The response is largely supported, but one measurement channel shows uncertainty. Review before acting. The system tells you which dimension failed and why.
Escalate
Multiple dimensions reject.
The system cannot commit. Route to a qualified human — with a complete explanation of what the system measured and where it found gaps.
Refuse
Outside measured territory.
The query falls outside what the system has verified. The decoder will not speculate. This is the most honest answer a system can give.
01
The sorting is the product.
A system that processes ten thousand inputs and correctly assigns postures has done something a language model cannot: it has told you which outputs to trust, which to review, and which to route to an expert — without requiring a human to read every one.
02
The posture is a measurement, not a guess.
Most AI confidence scores describe historical accuracy. VERITY's posture describes this specific output — whether the geometric evidence for this particular response, to this particular query, in this particular context, supports commitment. It is a per-instance measurement, not a population statistic.
03
Refusing is a feature, not a failure.
A system that can only answer is dangerous — it has no way to signal when it is guessing or confabulating. Refuse is the signal. It means the system reached the boundary of what it has verified and stopped there, rather than continuing into speculation.
What that means

Built for the domains that
cannot accept "probably right."

Under Daubert v. Merrell Dow Pharmaceuticals, scientific evidence must satisfy four criteria for court admissibility. Large language models fail all four — structurally. VERITY was designed to satisfy each one.

Daubert FactorLLM SystemsVERITY
Testable & FalsifiableNon-deterministic. Identical inputs produce different outputs. The methodology cannot be tested because there is no stable methodology to test.Geometric measurement is deterministic. Identical inputs produce identical results. The methodology is specified, stable, and reproducible.
Peer ReviewArchitecture is published. Outputs are not auditable. Attention weights do not explain specific claims in any form subject to meaningful review.Methodology is documented. The Semantic Fidelity Benchmark is public. Results are reproducible by any party with access to the same inputs.
Known Error RateHallucination rates vary by prompt, domain, and phrasing. The error rate for any specific output is unknown at the time of production.Every response carries a posture. The error rate is explicitly stated per category. Per-dimension decomposition is available for every output.
General AcceptanceNot accepted as a forensic evidence methodology by any professional forensic body.Geometric measurement of semantic content is established mathematics, independent of the ongoing controversy over AI generation.
Forensic AI Liability Series →
The principle

"An accurate but non-credible system is dangerous. It is right most of the time — so people trust it. But it is equally confident when right and when wrong. The errors are undetectable until harm occurs."

— Fidelity as a Measurable Quantity, Credasis AI