INDEX

Explanations

words ending in 'ed' or 'ing'

The neuron strongly activates on uncommon, content‐bearing roots (rare or specialized nouns and verbs) rather than on common function words.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 have

-2.28

in

-2.05

at

-1.75

can

-1.69

if

-1.59

all

-1.51

 before

-1.48

 could

-1.44

/**

-1.40

 distinctly

-1.38

POSITIVE LOGITS

 jego

1.54

 lentement

1.45

 kabát

1.41

 więks

1.41



1.40

 idiotic

1.38

 ayakkab

1.37

 bisiklet

1.35

 ürünleri

1.35

 irgend

1.32

Activations Density 0.101%