INDEX

Explanations

Some.

The neuron lights up on unusually frequent, content-bearing words in a document—i.e. the high-TF-IDF terms that characterize its main topic.

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

any

-1.70

any

-1.63

neries

-1.45

 principalement

-1.43

Any

-1.43

 anybody

-1.41

wasn

-1.41

 любой

-1.38

任何

-1.36

 qualsiasi

-1.32

POSITIVE LOGITS

 even

1.87

</h3>

1.83

 might

1.77

may

1.70

 still

1.66

 just

1.55

</h2>

1.50

なんですよ

1.48

 bahkan

1.47

}`}>

1.45

Activations Density 0.282%