INDEX

Explanations

code delimiters

The neuron fires on non‐alphanumeric symbol sequences commonly found in code—e.g. dollar‐sign identifiers (`$rootScope`), array indices (`:`), URL fragments (`://`), and hex color literals (`#f97e72`).

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

྅

-3.28

其妙

-2.91

 ſeveral

-2.52

 They

-2.45

 mädchen

-2.39

‪.‬‬

-2.34

⢕

-2.34

 mereka

-2.31

ビックリ

-2.23

 blancs

-2.19

POSITIVE LOGITS

er

3.17

2.88

ه

2.80

2.73

2.47

 Ersten

2.41

 they

2.38

We

2.36

甴

2.33

),

2.30

Activations Density 0.004%