INDEX

Explanations

but, or, and

This neuron flags unusually infrequent or “rare” tokens (i.e. high–information‐content words or punctuation like commas).

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

珢

-1.48

∅

-1.43

 egens

-1.38

bê

-1.34

%\]

-1.34

≅

-1.32

 devem

-1.30

 беже

-1.29

tê

-1.28

鹇

-1.27

POSITIVE LOGITS

and

2.31

or

1.84

but

1.74

 marvellous

1.34

 while

1.34

 questionable

1.29

jauh

1.25

 verbonden

1.22

 sizable

1.21

 gesteld

1.21

Activations Density 0.047%