INDEX

Explanations

code before semicolon or period

The neuron fires on tokens that are names of software libraries, frameworks, or code-generation tools (e.g. “protobuf,” “react,” “spark,” “yuitest,” 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

 their

-2.23

for

-2.11

at

-1.91

on

-1.91

 this

-1.87

 during

-1.83

 some

-1.80

by

-1.80

 each

-1.77

 from

-1.76

POSITIVE LOGITS

generally

1.56

 .......

1.56

usually

1.48

often

1.46

 …..

1.45

.(

1.35

could

1.34

 ........

1.34

 gigantic

1.33

 occuper

1.32

Activations Density 0.001%