INDEX

Explanations

constructor calls

The neuron fires strongly on code tokens that are class/type names (i.e. identifiers in UpperCamelCase).

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

ῆς

-1.02

actually

-1.01

শে

-0.99



-0.94

rte

-0.93

 bijzondere

-0.92

 gemeenten

-0.92

 médicament

-0.91

 contando

-0.90

ӳ

-0.89

POSITIVE LOGITS

2.05

1.57

()

1.48

(){}

1.40

 with

1.33

(){

1.30

():

1.11

 take

1.09

but

1.08

"></

1.07

Activations Density 0.011%