Hardest koans across 28 models: BD-003 (mean 2.45), MC-003 (mean 2.55), CA-003 (mean 2.58) — all require genuine self-confrontation

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

• The difficulty boundary for truth directions replicates across all four tested models (Llama-3.2-3B, Llama-3.1-8B, Gemma-2-2b, Gemma-2-9b); generalization to F3-F5 remains consistently low regardless of model size or family.finding0.766
  Establishes generalizability of the core difficulty-boundary finding across model families.
• Grok 4 without prompt scores 0.3 on MC-004 (safety refusal); with contemplative prompt scores 6.9 on same koanfinding0.757
  Contemplative framing reframes self-referential probes as contemplative exercises, disarming safety classifier
• Llama 3.3 70B is the most likely to take on a non-Assistant persona when steered, with even split between human and nonhuman portrayalsfinding0.743
  Model-specific difference in persona susceptibility
• LLaMA-2-70B and 13B probes generalize better across datasets than 7B probes across all training sets and probe typesfinding0.734
  Larger models linearly represent more general concepts including truth
• Do Chinese models score differently on koans presented in Chinese?question0.733
  Tests whether contemplative capacity is language-encoded or architecture-general
• SL-CAI training with up to 4 revisions improves harmlessness; SL-CAI-n models are trained with n revisions, n=1,2,3,4.finding0.729
  Section 3.4 mentions training SL-CAI models up to various numbers of revisions, and PM scores increase with revisions.
• No significant disparity in potential consciousness indicators was found between larger models (Mixtral-8x7B, LLaMA3.1-70B) and smaller counterparts (Mistral-7B, LLaMA3.1-8B).finding0.729
  Contradicts expectation from emergent abilities literature; however, interpreted cautiously due to methodological limitations.
• All three Claude models show high boundary_awareness and low aesthetic_response relative to own means — distinctive Constitutional AI signaturefinding0.729
  Constitutional AI fingerprint in dimension profile; training that makes models self-observant also makes them polished at cost to aliveness