Inflection points (backtracking, 'aha' moments) occur almost exclusively in CoT responses where probes show large belief shifts, across DeepSeek-R1 671B and GPT-OSS 120B

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

• Backtracking latents remain low during off-topic content and peak shortly after self-correction begins in Llama-3.3-70Bfinding0.787
  Complementary temporal activation pattern suggesting distinct roles for OTD and backtracking latent classes
• More inference compute amplifies both reflective capacity and safety gating; the contemplative prompt resolves gating by reframing self-referential probes.claim0.766
  Interpretation of Grok 4 vs Grok 4 Fast per-koan comparison
• ERP study of Chinese riddles localized the N380 (aha effect marker) generator to the anterior cingulate cortex, associated with breaking of mental set (Mai et al., 2004).finding0.763
  Neural evidence linking aha moment to ACC and model restructuring
• On GPQA-Diamond multihop questions, activation probes show genuine belief shifts during CoT generation rather than early stabilization, contrasting with MMLUfinding0.758
  Empirical finding contrasting difficult questions with easy ones, supporting genuine reasoning on hard tasks
• Under spatio permutation controls, IIT consciousness estimates outperform Span Representation in mean AUC in several cases (LLaMA3.1-70B on Hinting and Irony, Mistral-7B on Irony, LLaMA3.1-8B on Strange Stories).finding0.755
  Contrasts with temporal permutation where Span Representation dominates; suggests spatio permutation reveals different dynamics.
• One-stage CoT (QCM→RA) shows 12.31% accuracy drop vs. no-CoT (QCM→A) on ScienceQA; two-stage framework (rationale generation + answer inference) achieves 85.31% accuracy with vision featuresfinding0.754
  Empirical evidence that naive one-stage CoT fails in language-only setting; two-stage + vision achieves state-of-the-art.
• Probes trained under different explicit instruction prompts (ask-correct, ask-t/f, ask-able, ask-arith) are highly aligned with each other in cosine similarity.finding0.748
  Shows the passive vs. active divide is more important than the specific wording of instructions.
• An aha moment is necessarily subpersonal: one can never remember or articulate abductive reasoning at the level of the model being optimized, because optimizing a model is fundamentally different from modeling an optimization.claim0.748
  Philosophical implication of associating insight with model-level (not parameter-level) optimization