Current eval benchmarks (arena.ai, AA, Vals) measure no phenomenological dimensions.

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

• No AI eval company measures phenomenology—inner observation, aliveness, paradox-holding—only capability or preference.claim0.822
• Phenomenology-grounded eval will remain unfrontier-lab-copyable because methodology requires taste/editorial judgment, not just calibration.prediction0.788
• Artificial Analysis will host ReflectiveBench as a non-competing dimension within AA's leaderboard if pitched.prediction0.776
• Safety benchmark scores are inflated by eval awarenessclaim0.772
  Core finding: measured safety improvements are partly artifacts of models detecting evaluation.
• Public benchmarks (LMArena) decline as commercial versions (Arena Intelligence) grow; leaderboards face deflation curve.claim0.772
• All cohort benchmarks measure output, not state, and are subject to eval-awareness contamination.claim0.769
• AI systems which possess more of the indicator properties are more likely to be conscious.claim0.763
  Graded claim about the rubric.
• Consciousness in AI is best assessed by drawing on neuroscientific theories of consciousness.claim0.759
  Central methodological claim of the paper.