INDEX

Explanations

get, add, create, init

This neuron detects tokens that are (often capitalized or underscore-connected) function and API names in code.

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

now

-1.90

and

-1.80

 ‘’

-1.56

 зае

-1.35



-1.33

ὸ

-1.30

楽しめます

-1.28

⏑

-1.28

 шля

-1.21

 foremost

-1.20

POSITIVE LOGITS

Then

1.93

Using

1.82

Not

1.66

Even

1.64

1.53

While

1.52

Obviously

1.52

With

1.52

Perhaps

1.52

There

1.50

Activations Density 0.014%