INDEX

Explanations

words like "is," "to," and "be," often in conjunction with other words, indicating that it is identifying statements or relationships

anything

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

Negative Logits

Примечания

-0.53

 surla

-0.51

psack

-0.51

receptor

-0.49

asaran

-0.48

зулта

-0.48

requipa

-0.48

 حالی

-0.46

transQ

-0.46

꿔

-0.44

POSITIVE LOGITS

 anything

2.91

anything

2.61

 Anything

2.55

Anything

2.48

 ANYTHING

1.98

 anywhere

1.66

anywhere

1.51

 Anywhere

1.38

 aught

1.38

Anywhere

1.26

Activations Density 3.609%