INDEX

Explanations

end of phrases

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 alone

-0.10

 divert

-0.10

 delaying

-0.09

 Hein

-0.09

 Hans

-0.09

 mechanism

-0.09

 detailed

-0.08

 sift

-0.08

 Espresso

-0.08

 sophisticated

-0.08

POSITIVE LOGITS

 unused

0.14

 Unused

0.12

 useless

0.10

 clen

0.10

 Epid

0.10

 unnatural

0.10

Dyn

0.10

 UNUSED

0.10

 fists

0.10

 boca

0.10

Activations Density 0.010%

No Known Activations