INDEX

Explanations

three or respectively

The neuron fires on the structural markers of procedural steps (e.g. “comprises,” “sub-steps,” colons and “of”)—in other words, the language used to enumerate method steps.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 другой

0.77

 outro

0.75

 muchísimo

0.73

 വളരെ

0.72

 уровнем

0.70

 மிகவும்

0.70

超過

0.68

は約

0.68

ровень

0.67

 другое

0.66

POSITIVE LOGITS

分别是

1.52

分别为

1.40

 각각

1.39

三个

1.37

それぞれ

1.34

分别

1.27

 क्रमश

1.27

three

1.25

respectively

1.25

分別

1.24

Activations Density 2.689%