INDEX

Explanations

email addresses and urls

The neuron fires on domain name components and symbols in URLs and email addresses (e.g. “@”, “.com”, “.edu”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 pierwszej

-1.65

 which

-1.47

 primeiras

-1.45

໊

-1.43

 sirvió

-1.43

 drugie

-1.40

レ

-1.40

 pudiera

-1.39

 podría

-1.38

 プロ

-1.38

POSITIVE LOGITS

</strong>

3.38

</i>

3.13

</em>

3.08

1.97

？！”

1.73

if

1.67

 コイン

1.66

1.56

</h3>

1.56

 такі

1.55

Activations Density 0.015%