INDEX

Explanations

list item markers and section titles

The neuron strongly activates on non-letter tokens—punctuation marks, symbols, and numeric strings—i.e. formatting/UI elements rather than ordinary words.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 perme

1.24

ли

1.17

1.14

listView

1.13

("-

1.09

compat

1.05

жения

1.05

tics

1.05

tint

1.04

よりも

1.03

POSITIVE LOGITS

1.73

1.69

1.68

 February

1.66

1.65

1.63

Vgl

1.62

1.59

1.57

 October

1.56

Activations Density 3.046%