INDEX

Explanations

numbers and variables in equations

The neuron activates on numeric tokens and punctuation (digits, periods, hyphens), i.e. it picks out numbers (years, decimals, case/episode numbers, etc.).

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

is

-1.28

<<<<<<<

-1.23

bravo

-1.15

 what

-1.15

 dość

-1.14

 доволь

-1.13

 genaue

-1.13

 pouze

-1.12

 hivyo

-1.10

''

-1.09

POSITIVE LOGITS

巔

1.43

ofas

1.39

깽

1.34

olate

1.33

1.32

Koordin

1.28

1.27

1.26

 dekoratif

1.25

鸕

1.24

Activations Density 0.100%