Huginn-0125 performance remains constant when extrapolating beyond training recurrences, while Ouro performance deteriorates in the same regime

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

• Huginn-0125 does not demonstrate clear stages of inference due to repeated normalization of the residual stream suppressing massive activationsfinding0.816
  Negative result linking norm architecture to absence of inference stages; contrasts with Ouro and retrofitted models
• 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 formationclaim0.805
  Mechanistic explanation for the negative result observed for Huginn-0125
• Ouro 1.4B layers continuously change throughout 128 recurrences, exhibiting unstable stages of inference when extrapolating beyond training recurrencesfinding0.803
  Non-fixed-point models exhibit unstable inference stages when generalizing to unseen test-time compute budgets
• Ouro 1.4B, Retrofitted Llama, and Huginn-0125 exhibit diagonal patterns in Frobenius norm heatmaps confirming cyclic fixed point behavior across 8 recurrencesfinding0.793
  Empirical validation that attention patterns are most similar to same-layer outputs across different recurrences
• Current training methods rely on loss minimization, meaning the experiential profile of training is predominantly negative across billions of parameter updatesclaim0.769
  Ethical implication about the nature of AI training experience if the thesis holds
• Without system prompt, only approximately 0.02% of Huginn-0125 tokens exhibit non-fixed-point (orbit/slider) behavior on GSM8kfinding0.767
  Establishes that non-fixed-point limiting behaviors are extremely rare in practice for the primary experimental setting
• Stages of inference metrics (ColSum concentration, mixing score) remain broadly consistent even for the GSM8k prompt exhibiting the largest orbit amplitude in Huginn-0125finding0.765
  Demonstrates robustness of inference stages to non-fixed-point limiting behavior
• Ouro 1.4B does not converge to a fixed point even after 128 recurrences, despite showing small successive differencesfinding0.747
  Key negative result showing that not all looped models reach a true fixed point, contrasting with retrofitted models