INDEX

Explanations

not followed by another word

The main thing this neuron does is detect occurrences of the word “not.”

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

戗

-2.48

떳

-2.47

are

-2.47

賔

-2.33

粜

-2.31

쑹

-2.28

 anderen

-2.25

 swells

-2.23

鼱

-2.23

횃

-2.22

POSITIVE LOGITS

 necessarily

2.83

2.63



2.38

 котором

2.25

𝐲

2.23

囌

2.17

亊

2.16

 テラス

2.16

ching

2.14

 analogous

2.11

Activations Density 0.189%