INDEX

Explanations

among the

The neuron fires on uncommon, longer content words—i.e. rare or domain-specific tokens rather than common short words.

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

 that

-2.70

 trattamento

-1.68

 możesz

-1.62

 男の子

-1.56

 ジュニア

-1.55

 ACTUALLY

-1.51

 شروع

-1.48

 ครับ

-1.47

 赤ちゃん

-1.46

 cercano

-1.45

POSITIVE LOGITS

for

2.64

of

1.93

to

1.64

НУ

1.59

如

1.58

antism

1.47

 видит

1.47

 птица

1.47

icoot

1.47

1.46

Activations Density 0.048%