concept
active
concept:factualityFactuality
Scoped definition of 'truth' used in the paper: the truth or falsehood of declarative factual statements
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Christiano et al. (2021) framework motivating the problem of determining whether a model 'believes' a statement; cited as core motivation
Concepts (1)
concept
- Truth Direction in LLM Latent Spaceassociated_withA specific direction in an LLM's residual stream that encodes the truth or falsehood of factual statements
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- The representation of something currently being the case; a variable feature dimension of conscious contents distinguishing ideas from hallucinations of factuality
- A correctness condition requiring assertions to be true.
- The capacity to have beliefs, desires, intentions; discussed in the context of AI and speech acts.
- Alexander's judgment on blueprint-driven design assumption; you cannot make a human or daffodil from detailed specification.
- The directness of motivation by practical concerns, characteristic of living processes in the examples.
- The paper's operationalization of truthfulness as simple, unambiguous propositional statements that can be labeled true or false
- Formal notion of what constitutes an individual agent; bridges Buddhist and information-theoretic perspectives.
- The sequential, continuous order of text, often challenged by diagrammatic branching.