INDEX

Explanations

translated as

This neuron is tuned to detect words and phrases related to translations (e.g., “translated,” “translation,” “переводится,” etc.).

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



-1.49

 informasjon

-1.48

鋮

-1.43

ufe

-1.41

nadequate

-1.41

菈

-1.41

<bos>

-1.41

 saken

-1.39

 ferdig

-1.38

POSITIVE LOGITS

 outras

1.66

“

1.62

the

1.52

 called

1.48

 otras

1.48

 сейчас

1.40

1.35

 Denken

1.33

 есть

1.31

 Resultat

1.26

Activations Density 0.025%