INDEX

Explanations

roles of people

The neuron strongly activates on words naming people or agents (e.g. customer, buyer, reader, user, patient, etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 were

-1.80

</b>

-1.76

</strong>

-1.70

</h4>

-1.64

 they

-1.49

 podrían

-1.48

.”

-1.48

denver

-1.46

</em>

-1.43

).

-1.43

POSITIVE LOGITS

 himself

2.14

if

1.96

who

1.69

他

1.65

 wants

1.63

or

1.62

 gets

1.62

如果

1.61

animity

1.58

ನ

1.56

Activations Density 0.128%