Nocturnal sleep doubled the prevalence of insight-dependent performance improvements on a stimulus-response sequence task (Wagner et al., 2004).

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

• Simulated agent achieves perfect performance after trial 12 following Bayesian model reduction (sleep), versus ~14 trials without BMR.finding0.755
  Key result showing BMR dramatically accelerates rule learning in simulation
• We hypothesize that insight produces a profound reduction in the latency of evoked neuronal responses when subjects know the meaning of cues (have learned a rule), equivalently an increase in ERP amplitude to initial cues in a sequence.hypothesis0.744
  Empirically testable neural prediction of the active inference model of insight
• Tononi & Cirelli (2006) — Sleep function and synaptic homeostasisconcept0.735
  Synaptic homeostasis hypothesis cited as neurobiological analog of non-REM Bayesian model reduction
• Representational dynamics of causal emergence align with reward improvement in most tasks.finding0.731
  The trajectory of causal emergence through training mirrors the reward improvement curve across the majority of tested environments.
• 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 benchmarkshypothesis0.729
  Comparative prediction motivating future work contrasting different approaches to LLM self-knowledge
• Monosemanticity and entanglement of SAE features were benchmarked for clinical taxonomy grounding across SleepFM, REVE, LaBraM.finding0.727
  Quantitative assessment of feature quality using clinical concepts across models.
• fMRI showed increased right hemisphere anterior superior temporal gyrus activity for insight solutions, with EEG showing a gamma-band burst in the same region beginning 0.3s prior to insight (Jung-Beeman et al., 2004).finding0.727
  Neural correlate of insight used to support prediction about early neural activity following structure learning
• Trainable intervention (DAS) finds sparser gender representations than linear probing, suggesting probing overestimates causal coverageclaim0.726
  Interpretive claim from Case Study II about the distinction between correlational probes and causal interventions