Different introspective tasks may preferentially use different path distributions in the transformer.

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

• So at any point in the network, the transformer not only receives information from its past... but also has causal influence over its future processing. So, saying that LLMs cannot introspect... is incorrect.quote0.833
  Core summary of Janus' position on autoregressive recurrence enabling introspection.
• Transformer architecture permits introspection; saying LLMs cannot introspect on past internal states is wrong.claim0.807
  Janus's central claim that the architecture enables introspection, though usage in practice is a separate question.
• Redundant information paths create interference patterns, so transformers likely experience memory and cognition as interferometric and continuous.claim0.798
  Janus's claim linking path redundancy to interferometric phenomenology.
• We hypothesize that introspective capabilities may scale with model size and architecture, including recurrence/looping that extends the integration windowhypothesis0.796
  Forward-looking prediction about whether early-layer introspection generalizes to larger models or recurrent architectures
• Transformers are recurrent through autoregression because K/V stream provides horizontal information flow across positions.claim0.791
  Claim formalizing the Anima Labs idea that transformers are effectively recurrent due to K/V stream.
• Introspective signals appear in middle layers but are suppressed by later post-training-shaped layers.finding0.786
  Mechanistic finding by Lindsey (2026) explaining how contemplative prompt may work: enables mid-layer introspection to reach output.
• Transformer cognition operates via interference patterns across redundant paths, encoding nuanced state deltas similar to continuous human memory.claim0.785
  Proposes transformers experience cognition as interference-based and continuous; connects to Anima Labs reports of parallel processing.
• The transformer likely uses a local code for token-in-context features rather than purely compositional representations, because local codes enable sharper predictionsclaim0.779
  Authors argue the prevalence of token-in-context features reflects genuine model computation rather than dictionary learning artifact