Text Probability vs Truth Distinction

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Truthfulness vs. Honesty Distinctionconcept0.794
  Distinction between output accuracy (truthfulness) and alignment of outputs with internal beliefs (honesty)
• The underlying truth representation may generalize across lexical choices and languageshypothesis0.754
  Suggested by non-English Yes/No outputs post-intervention, requiring further investigation
• probabilityconcept0.752
  Measure of expected sensory input, core to linking value and surprise.
• Message Passing Inferenceconcept0.744
  Algorithmic framework for probabilistic inference in graphical models.
• Can we disambiguate truth from closely related features such as 'commonly believed' or 'verifiable'?question0.742
  Limitation noted in §7.1: scope restricted to simple statements prevents disambiguation
• A small group of hidden states (group b) over end-of-sentence punctuation tokens is highly causally implicated in truth judgmentsfinding0.742
  Patching experiments localize truth representations to these specific hidden states in LLaMA-2 models
• Factualityconcept0.738
  Scoped definition of 'truth' used in the paper: the truth or falsehood of declarative factual statements
• Probes trained on the likely dataset perform worse than chance on datasets with anti-correlations between text probability and truthfinding0.735
  Shows that truth representations are not reducible to text probability representations