INDEX

Explanations

Pdate, Pool, Fade

The neuron consistently fires on single-character tokens (individual letters used as identifiers or symbols).

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

 chiesa

-1.39

 alcune

-1.38

或者

-1.38

你会

-1.34

 alcuni

-1.34

私の

-1.33

Sede

-1.32

妡

-1.31

IFYING

-1.30

ایک

-1.28

POSITIVE LOGITS

1.80

to

1.68

 both

1.57

by

1.45

 that

1.38

1.34

 Both

1.30

 zowel

1.30

 Though

1.23

Both

1.23

Activations Density 0.077%