INDEX

Explanations

Q names

The neuron consistently responds to the uppercase letter “Q” when it appears as its own token (e.g. in names or code-like strings).

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

 This

-1.63

 ikke

-1.52

rania

-1.44

шего

-1.42

 هذا

-1.40

 ersche

-1.39

 bekämp

-1.38

ного

-1.36

 này

-1.29

şi

-1.28

POSITIVE LOGITS

釹

1.41

婀

1.39

 then

1.38

不好的

1.33

AR

1.31

AL

1.30

選んだ

1.30

一旁

1.27

AN

1.26

Activations Density 0.058%