INDEX

Explanations

<?php or <?xml code

The neuron strongly activates on the "<?" token (the opening tag for PHP or XML), effectively detecting code/script openings.

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

<bos>

-1.59

鲽

-1.41

 fundamenta

-1.37

 知道

-1.34

覚悟

-1.34

そして

-1.32

apartament

-1.30

as

-1.29

谁

-1.27

だった

-1.26

POSITIVE LOGITS

為に

1.52

1.46

ini

1.46

際には

1.43

すっ

1.42

et

1.41

つづく

1.40

1.35

gelegt

1.34

1.33

Activations Density 0.003%