Do other high-level operators such as reasoning style or metacognitive framing likewise localize to specific attention heads?

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

• Attention heads with positive projection on reflection direction are sparse and located mostly in deeper layers of DeepSeek-R1-Qwen-1.5Bfinding0.780
  Structural finding about which attention heads control reflection behavior
• Some attention heads partially specialize in copying for words that split into two tokens without a space prefix, attending from fragmented token to complete tokenfinding0.775
  Interesting special case of copying behavior related to tokenization artifacts; primitive precursor to induction heads
• Attention algorithms are usually distributed across attention headsclaim0.772
  Claim supported by VPD's recovery of cross-head attention subcomponents, noted in footnote.
• MLPs primarily serve token-wise processing and are closely associated with knowledge storage, making attention heads the appropriate locus for sequence-level style modulationclaim0.772
  Architectural rationale for why Style Modulation Heads are in attention layers rather than MLPs
• Similar superposition phenomena may exist in self-attention layers and similar sparse autoencoder methods may extract useful structure from attentionhypothesis0.764
  Extension of superposition hypothesis to attention layers as future research direction
• One-layer model attention heads encode Python-specific skip-trigrams including indentation-based elif/else prediction and function signature patternsfinding0.758
  Concrete example from examining expanded QK/OV matrices showing how specific programming language structure is encoded in attention weights
• All 32 attention heads at layer 3 achieve 100% localization accuracy for injections at layer 2 (5-way classification, 20% chance)finding0.757
  Striking mechanistic finding that injection creates universally detectable perturbation in residual stream immediately downstream
• Most attention heads in one-layer models dedicate an enormous fraction of their capacity to copying behaviorclaim0.757
  Empirical observation from examining expanded OV/QK matrices; approximately 10 out of 12 heads show significant copying