INDEX

Explanations

introducing possibilities

The neuron detects hedge or speculation language that qualifies or softens statements.

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

are

-1.65

擽

-1.59

 disponibile

-1.59

 принцип

-1.48

)”.

-1.47

ance

-1.41

 送料

-1.39

nologies

-1.38

 görmek

-1.37

or

-1.36

POSITIVE LOGITS

 sogar

1.80

 μας

1.56

 even

1.49

 الأولى

1.48

 termed

1.40

 addirittura

1.38

dáms

1.36

 البعض

1.34

 jsme

1.34

äure

1.34

Activations Density 0.015%