Boolean neurons not fitting a specific sub-type.
What It Does
Boolean.General neurons carry boolean-like signals that resist clean classification into the six specific sub-types. They often appear at the intersection of multiple Boolean categories — a statement that is both negative and comparative, or both existential and affirmed. They also encode implicit boolean logic that emerges from context rather than explicit boolean markers.
How It Behaves
As the largest Boolean sub-type by count, General neurons serve as a catch-all for the semantic complexity of boolean reasoning that cannot be reduced to a single pure signal. They tend to appear throughout all layers, reflecting their general-purpose role. When specialized Boolean neurons are suppressed (through GI Surgeon edits or natural model variation), General neurons sometimes compensate — a form of representational redundancy.
Research Example
In cross-model analysis, Boolean.General neurons show the most stable distribution across all 11 models tested — they are present in consistent proportions regardless of model architecture or size. This suggests they encode something fundamental about how boolean reasoning is represented in transformers, independent of architectural choices.