Huginn-0125 fails to develop stages of inference because its repeated residual stream normalization prevents the growth in activation magnitude required for compression behavior and sink formation

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

• Huginn-0125 performance remains constant when extrapolating beyond training recurrences, while Ouro performance deteriorates in the same regimefinding0.805
  Correlates stable fixed-point behavior with out-of-domain generalization performance at test-time
• Stages of inference metrics (ColSum concentration, mixing score) remain broadly consistent even for the GSM8k prompt exhibiting the largest orbit amplitude in Huginn-0125finding0.793
  Demonstrates robustness of inference stages to non-fixed-point limiting behavior
• Without system prompt, only approximately 0.02% of Huginn-0125 tokens exhibit non-fixed-point (orbit/slider) behavior on GSM8kfinding0.773
  Establishes that non-fixed-point limiting behaviors are extremely rare in practice for the primary experimental setting
• Learning the likelihood matrix A via Dirichlet parameter accumulation is formally identical to associative or Hebbian plasticity.claim0.749
  Connection between learning rule and synaptic plasticity.
• Patching h[1] with a divergent representation can activate distinct, hidden pathways that result in misleadingly confirmatory behavior and/or undetected behavior.quote0.744
  Load-bearing description of the core pernicious divergence mechanism illustrated in Figure 1
• The middle layer residual stream features are causally implicated in multi-step reasoning.claim0.744
  Features for Kobe Bryant, California, Lakers participate in computing the capital answer.
• 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.claim0.743
  Key theoretical claim linking active inference to physics in Section 2.
• The likelihood of a dedicated feature for a concept (element, city, animal, food) follows a sigmoid in log-frequency of the concept in training data, with threshold frequency inversely proportional to number of alive features.finding0.742
  Quantitative relationship between concept frequency and feature presence.