Large Language Models Understand Language

Claim: Large language models genuinely understand and comprehend human language rather than merely predicting tokens through statistical pattern matching.

MOSTLY SOLID Worldwide#ai #llm #machine-learning

Compare

Score explainer

What's behind the number?

28.0/ 100

Low BS signals

BSR = 100 × E × (0.60 + 0.25 × D + 0.15 × H)

= 100 × 0.381 × (0.60 + 0.25 × 0.443 + 0.15 × 0.167) = 28.0

Formula specificationPublishedMethodology

E · 0.60

Epistemic falsity

38.1%

How detached the claim is from physical reality and evidence.

D · 0.25

Defensive armor

44.3%

Ad-hoc excuses, vague language, dogma and operator dependence.

H · 0.15

Harm & grift

16.7%

Medical risk, money drain, conspiracy and cult dynamics.

Calculation transparency

Formula specificationPublished
Weights and transformationsPublished
Ranking eligibilityEligible

What the calculation exposes

01 · Input mapping

Each answer has a published weight from 0 to 1.

02 · Calculation rule

BSR = 100 × E × (0.60 + 0.25 × D + 0.15 × H)

03 · Result provenance

liquid/d1-ultra-20261015 · Oct 30, 2026

Model structure

10decision dimensions
46response categories
Separate confidenceEpistemic confidence is computed from the weight variance, not the model's own field.

Epistemic confidence: 36.0%

Search interest is displayed separately. It does not affect the score. Inspect the decision model