INDEX

Explanations

file extensions and tags

The neuron strongly activates on standalone tokens that look like file‐ or domain‐name extensions and header/macro suffixes (e.g. “com”, “jar”, “md5”, “h”, “html”, “de”, or “_H”).

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

']));

-1.55

</i>

-1.52

 рисунок

-1.45

࿚

-1.45

ATION

-1.45

co

-1.44

！

-1.44

</u>

-1.43

奶油

-1.41

鮪

-1.41

POSITIVE LOGITS

の影響

1.70

撂

1.63

alpine

1.59

tyimages

1.56

metido

1.50

 gezondheid

1.49

都想

1.45

熄

1.45

 jedes

1.45

1.44

Activations Density 0.006%