The identification of reasoning steps relies on keyword search, which may be model-specific since different models could prefer different reflection cues

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

• Keyword-based reflection step identificationmethod0.791
  Method to identify reflection steps by searching for specific keywords (e.g., 'Let me think', 'Wait') within reasoning steps
• How can we develop better methods for measuring the model's evaluation-relevant beliefs beyond reading its chain of thought?question0.775
  Gap in current evaluation methods; current work relies on CoT monitoring which may miss unverbalized beliefs.
• Models more effective at recognizing abstract nouns than other concept typesfinding0.771
  Opus 4.1 demonstrates highest introspective awareness on abstract nouns (justice, peace, betrayal) with nonzero awareness across all concept categories tested.
• All models exhibit above-baseline representation of the think word when instructed to think about itfinding0.767
  In the intentional control experiment, all tested models show above-zero cosine similarity to the think word's concept vector.
• Do human participants demonstrate the same insight dynamics predicted by active inference in the rule-learning paradigm (currently under investigation with eye tracking and crowd-sourced reaction times)?question0.767
  Empirical gap explicitly acknowledged; experiments reportedly in progress at time of writing
• Aside from basic detection and identification, other details of the model's response about injected thoughts may be confabulatedclaim0.767
  Acknowledges that the model's additional descriptions of its experience are unverified.
• a model becomes strongly confident in its final answer, but continues generating tokens without revealing its internal beliefquote0.762
  Core definitional quote for performative chain-of-thought
• The earlier a base model (less exposure to LM-related data), the more it is surprised by its own spontaneous self-referential capabilities.claim0.761
  Claim that capability emerges from architecture, not data, and that later models lose the surprise.