Internal states appear to encode Bayesian beliefs about hidden external states.

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

• If internal states encode a probability density over external states, then it should be possible to predict external states from internal states.hypothesis0.882
  The testable hypothesis driving the active inference analysis in the simulation.
• We hypothesize that group (b) hidden states store a representation of the statement's truthhypothesis0.809
  Motivating hypothesis driving the remainder of the paper's analysis after patching localization
• Internal states significantly predicted motion of external subsystems; best prediction for the farthest subsystem (magenta circle, Fig 4d).finding0.780
  Result of canonical variates analysis showing statistical dependency between internal states and external motion.
• Any system minimizing free energy will appear to engage in implicit Bayesian inference of hidden external causes.hypothesis0.772
  Predicts that internal states encode posterior beliefs about external world through gradient descent on free energy.
• A small group of causally-implicated hidden states encodes LLM truth representations, localized over clause-ending punctuation tokensclaim0.765
  Localization result from patching experiments; identifies group (b) hidden states as the locus of truth representations
• Under the Markov blanket assumption together with NESS, a generalised synchrony appears, such that the dynamics of internal states can be cast as performing inference over external states via minimisation of variational free energy.claim0.760
  Key theoretical claim linking active inference to physics in Section 2.
• Intentional Control of Internal Statesfinding0.756
  Models can modulate their internal representations when instructed or incentivized to 'think about' a concept; effect replicates across all tested models regardless of capability.
• We hypothesize that emotion states are more persistent because they correspond to genuinely stateful internal representations, not merely local surface contenthypothesis0.754
  Proposed explanation for why emotion probes are more persistent than variance-matched random probes