INDEX

Explanations

case statements

The neuron fires on programming‐related tokens—things like control-flow keywords (e.g. “case”, “break”), event constants (“ACTION_DOWN”), tag names (“li”), and similar code symbols.

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

not

-1.64

</b>

-1.57

Präsident

-1.46

is

-1.37

 meant

-1.27

 abundantly

-1.27

 ungefähr

-1.26

</strong>

-1.22

 them

-1.22

 example

-1.21

POSITIVE LOGITS

錞

1.50

uito

1.44

 مانند

1.35

 otros

1.34

 koncentr

1.33

 іде

1.27

 produisent

1.27

 beliebt

1.26

してくれます

1.24

 from

1.23

Activations Density 0.002%