Reaction
Could monosemantic or monofunctional MLP neurons limit how many entities a model knows?
A user asks whether, if MLP neurons each have one meaning or function, a model’s neuron count sets that limit.
TLDR
A user asks whether the number of entities a model knows would be bounded by its neuron count if MLP neurons are monosemantic or monofunctional. The post frames both the premise and the proposed limit as questions, not findings.
Combined views
4.3K
2 Sources, first seen ago
