INDEX

Explanations

programming identifiers

The neuron fires on code identifiers that carry numeric suffixes (e.g. the base token of “num1,” “first2,” “len1,” “color1,” 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

 with

-1.47

 after

-1.38

or

-1.28

しましたが

-1.22

takkan

-1.17

 from

-1.17

at

-1.17

得不

-1.09

and

-1.05

in

-1.03

POSITIVE LOGITS

1.19

 beantragt

1.09

必要がある

1.07

 alcune

1.06

atele

1.05

 جون

1.02

 mengembangkan

1.02

 "'";

1.02

 qualche

0.99

羮

0.98

Activations Density 0.008%