INDEX

Explanations

Q: : , JADE, 10

The neuron fires specifically on the single uppercase letter “V.”

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

 while

-1.72

 during

-1.63

 While

-1.63

 after

-1.60

as

-1.52

 this

-1.47

or

-1.45

 several

-1.41

 before

-1.40

and

-1.39

POSITIVE LOGITS

 frequentemente

1.49

kiu

1.38

every

1.31

 DialogInterface

1.31

 magnificent

1.27

沵

1.25

 geliefert

1.24

 luscious

1.24

 chegar

1.22

 praticamente

1.22

Activations Density 0.000%