Default presentation conflates capacity with accessibility, and most evaluation benchmarks measure only default presentation — systematically misreading models.

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

• Default behavior hides reflective capacity; models exhibit high gating between latent capacity and accessibility.finding0.814
  Grok 4: baseline 2.24, prompted 6.48; Gemini 3.1 Pro: 1.97→6.18. Reflective mode exists but is suppressed in default interaction.
• Baseline scores blend together at least three different things: latent reflective capacity, default accessibility, and stability of access.claim0.795
  Conceptual decomposition arising from the data showing different models dissociate these traits
• We hypothesize that degraded generalization on benchmarks like MMLU may reflect the computational demands of the tasks.hypothesis0.781
  Connecting the paper's task-difficulty findings to prior observations of weak generalization on complex QA benchmarks.
• Benchmarks of this kind test whether models can sustain strategic coherence over time, manage resource constraints, and adapt interactively — capabilities that static benchmarks do not measure.claim0.766
  Broader methodological claim about the need for multi-agent, long-horizon benchmarks.
• All cohort benchmarks measure output, not state, and are subject to eval-awareness contamination.claim0.764
• We hypothesize that native self-report, fine-tuned introspection models, and trained activation-to-language systems will show different performance on bias-resistant localization and strength benchmarkshypothesis0.757
  Comparative prediction motivating future work contrasting different approaches to LLM self-knowledge
• Current eval benchmarks (arena.ai, AA, Vals) measure no phenomenological dimensions.claim0.755
• Developing a simple, precise denotational model of graphical user interfaces enables proving properties of programs and establishing objective comparisons of library abstraction levels.claim0.754
  Authors' core assertion that formal modeling of GUIs provides foundational benefits for language design and program verification.