INDEX

Explanations

don't or haven't

The neuron strongly activates on negative contraction prefixes (e.g. the “don”/“haven” in “don’t,” “haven’t,” etc.), i.e. words expressing negation.

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

it

-1.88

 when

-1.56

tedì

-1.52

]=>

-1.52

ヵ

-1.47

でもあります

-1.46

 sudadera

-1.45



-1.42

 chaqueta

-1.42

 basada

-1.41

POSITIVE LOGITS

("")){

1.60

gangss

1.60

ꠥ

1.56

uksia

1.48

$};

1.46

 kurie

1.45

 Если

1.44

 не

1.41

 hushed

1.38

 dabar

1.38

Activations Density 0.038%