INDEX

Explanations

closing punctuation and code

The neuron fires on closing code‐syntax tokens—like “)”, “}”, “>”, and “;”—i.e. characters that mark the end of programming or markup constructs.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

—

-1.93

și

-1.87

ABASES

-1.64

ン

-1.58

、

-1.42

 such

-1.41

ácidos

-1.38

ты

-1.36

 такой

-1.35

và

-1.34

POSITIVE LOGITS

’

1.84

</b>

1.59

ardez

1.49

 übrigens

1.48

จริงๆ

1.43

AfterViewInit

1.38

дження

1.35

 Actually

1.34

 Probably

1.30

 сча

1.30

Activations Density 0.065%