INDEX

Explanations

both X and Y contrast

The neuron fires on categorical adjectives that set up binary oppositions (e.g. direct / indirect, public / private, small / large, positive / negative).

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

and

-1.96

niająca

-1.48

므

-1.39

 diensten

-1.37

 что

-1.34

 neither

-1.31

 artikelen

-1.30

感觉到

-1.27

jy

-1.26

regation

-1.25

POSITIVE LOGITS

 many

1.58

opheles

1.57

 those

1.52

 juos

1.52

</h5>

1.52

 låter

1.45

رير

1.44

罅

1.43

 krä

1.40

 handlar

1.39

Activations Density 0.044%