INDEX

Explanations

negation followed by words

The neuron fires on simple negation tokens—words like “no,” “nothing,” “none,” (and their equivalents in other languages).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

！（

-1.66

-1.62

-1.52

For

-1.41

 pentru

-1.40

每个

-1.39

也是

-1.38

if

-1.38

 برای

-1.38

ห

-1.36

POSITIVE LOGITS

of

2.28

 verpflichtet

1.59

ffine

1.55

 wiederhol

1.52

 piece

1.50

'),

1.50

弾

1.49

 other

1.49

 erweitert

1.47

 feitos

1.43

Activations Density 0.007%