INDEX

Explanations

United States

The neuron fires on verbs and phrases that establish a relation between entities—especially causal or possession‐type links (e.g. “causes,” “related,” “give,” “have four”).

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

Generally

-1.58

Those

-1.54

Numerous

-1.52

Usually

-1.49

Honestly

-1.49

Typically

-1.48

 princesas

-1.42

 Kriege

-1.42

Sometimes

-1.41

These

-1.40

POSITIVE LOGITS

of

2.39

 also

2.34

if

1.92

you

1.92

 they

1.84

 each

1.77

 even

1.63

 there

1.62

 ajoute

1.55

而

1.53

Activations Density 0.005%