INDEX

Explanations

hexadecimal strings

The neuron spotlights sequences of hexadecimal characters (i.e. hex‐encoded IDs or hex dumps).

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

 like

-1.44

 time

-1.40

 povez

-1.39

 better

-1.35

 from

-1.32

 those

-1.31

 relatively

-1.28

 both

-1.27

 fewer

-1.23

and

-1.23

POSITIVE LOGITS

 hatta

1.81

of

1.77

 semblent

1.46

()){

1.41

 tranſ

1.38

 datter

1.38

 miele

1.37

 œufs

1.35

Faça

1.34

 meren

1.34

Activations Density 0.041%