INDEX

Explanations

endpoints, themes, peers, train, DEMO, img, file, kill, frame, SIM, atom, email

This neuron activates on code identifiers that include underscores (e.g. snake_case or macro names).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 sebagainya

-1.57

€“

-1.46

//

-1.42

⢪

-1.41

The

-1.37

etc

-1.35

 ۱۳

-1.33

);

-1.31

 мои

-1.30

 plaît

-1.30

POSITIVE LOGITS

it

1.67

who

1.59

{~

1.57

 which

1.55

 there

1.41

our

1.38

 those

1.35

not

1.34

 information

1.34

 where

1.32

Activations Density 0.090%