INDEX

Explanations

code blocks or structures

The neuron fires on code tokens that introduce or name functions/methods—i.e. identifiers immediately followed by a parenthesized parameter list.

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

whose

-1.69

(«

-1.56

已经在

-1.45

whom

-1.41

‘‘

-1.38

»,

-1.37

又は

-1.37

«

-1.35

 "..\..\

-1.34

was

-1.32

POSITIVE LOGITS

one

1.70

 into

1.55

 from

1.48

It

1.47

as

1.45

 There

1.39

 there

1.38

 sogenannten

1.34

our

1.34

 entsprechenden

1.33

Activations Density 0.038%