INDEX

Explanations

value in things and others

The main thing this neuron does is spot generic placeholder nouns (e.g. “things,” “objects,” “others,” “fact”).

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

in

-2.09

 were

-1.68

let

-1.47

 need

-1.45

 needed

-1.42

 such

-1.41

 unless

-1.38

-1.35

for

-1.31

 then

-1.30

POSITIVE LOGITS

ᑎ

1.38

 mineures

1.35

 aucun

1.34

எ

1.30

 präsenti

1.30

 persec

1.27

 révélé

1.25

 disreg

1.23

 impra

1.21

ovala

1.20

Activations Density 0.078%