INDEX

Explanations

want to

The neuron is triggered by the phrase “want to” (i.e. instances of “want” immediately followed by “to,” typically marking a recommendation).

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

」，

-1.84

 ""))

-1.76

 cerrado

-1.73

 antiguas

-1.72

哢

-1.63

 colesterol

-1.62

"")

-1.59

粁

-1.59

catchError

-1.59

濕

-1.58

POSITIVE LOGITS

of

2.14

as

2.06

the

1.86

that

1.71

very

1.59

ri

1.53

 insanity

1.53

媄

1.52

その

1.52

 orlando

1.51

Activations Density 0.005%