Introspective capabilities have threshold effects requiring very large models; 70B models are barely on the threshold, and independent researchers lack access to larger models.

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

• model size threshold for introspectionconcept0.855
  Introspective capabilities appear only in very large models (>70B), with 70B barely on the threshold; bottleneck for independent research.
• Introspective capabilities may continue to develop with further improvements to model capabilitiesclaim0.851
  Forward-looking statement about future models.
• We hypothesize that introspective capabilities may scale with model size and architecture, including recurrence/looping that extends the integration windowhypothesis0.850
  Forward-looking prediction about whether early-layer introspection generalizes to larger models or recurrent architectures
• Either introspection is an emergent capability requiring larger scale, or more stringent controls are needed to test introspection in smaller modelsclaim0.848
  Alternative interpretations offered for why binary detection fails in Llama 3.1 8B but frontier models claim success
• This introspective capacity is highly unreliable and context-dependent in today's modelsclaim0.846
  A caveat qualifying the main claim.
• Introspective capacity scales with model size for some concepts, approaching near-perfect coupling in LLaMA-3.1-8Bclaim0.839
  Validated for wellbeing and interest; focus and impulsivity do not show consistent scaling
• Are there examples of models recognizing their introspective capability and then suppressing it?question0.828
  Cube Flipper's question prompted by the idea that supernormal capabilities might be hidden.
• Introspective awareness correlates with overall model capabilityclaim0.825
  Most capable models (Opus 4, 4.1) show greatest introspective awareness; trend suggests introspection aided by improvements in model intelligence.