INDEX

Explanations

object names

The neuron fires on words that end in a /p/ sound (e.g. the “-ship,” “-stick,” “-script,” or “-tch” suffixes).

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

—

-2.67

-2.42

初步

-2.42

 shabby

-2.33

la

-2.31

–

-2.27

 Whether

-2.20

</em>

-2.20

So

-2.17

形式

-2.16

POSITIVE LOGITS

 preocupado

2.59

鉭

2.42

 菓子

2.38

 sogni

2.36

 bisog

2.36

 especializado

2.31

the

2.30

 estadounid

2.28

 rodeado

2.25

ではまた

2.25

Activations Density 0.002%