claim
active
claim:at-least-in-simple-cases-llms-track-rigorous-definitions-of-belief-over-time-while-processing-free-form-textAt least in simple cases, LLMs track rigorous definitions of belief over time while processing free-form text
Overall conclusion.
Source paper
extracted_from(2026) · Eric Bigelow · Raphaël Sarfati · Daniel Wurgaft · Owen Lewis +4
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Central thesis.
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.
- We hypothesize that LLMs represent correctness of arithmetic expressions differently from factual statements.hypothesis0.820Core working hypothesis motivating the factual vs. arithmetic task split in the experimental design.
- Establishes that the observed linear structure is not merely a representation of text probability
- Conclusion from Experiment 2 on Leap-of-Thought.
- Sharma et al. result supporting cross-modal alignment: language-only models implicitly encode visual structure
- Chalmers' interpretivist framework adopted for Level 1 assessment.
- discussion of potential confounds
- The central research question motivating the paper
- Central empirical conclusion of the paper about the fundamental limits of truth directions.