INDEX

Explanations

speak about or with

The neuron fires on various forms of “speak” (including “speech” and “speaker”), i.e. tokens related to talking/speaking.

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

</b>

-3.06

–

-2.78

 slick

-2.63

ウォーター

-2.61

 encompass

-2.48

—

-2.47

籏

-2.36

-2.34

 pretend

-2.33

POSITIVE LOGITS

 giapponese

3.25

筚

3.14

 femminile

3.11

 内衣

3.05

ڧ

3.02

 estadounid

2.97

kemeja

2.88

sudadera

2.83

 特写

2.81

 аккура

2.81

Activations Density 0.017%