INDEX

Explanations

drastically different outcomes

New Auto-Interp

Configuration

Prompts (Dashboard)

238,145 prompts, 512 tokens each

Dataset (Dashboard)

lmsys + oasst1

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

主人公

0.39

 cili

0.37

主角

0.37

mir

0.36

 Rosen

0.36

effectively

0.36

 effectively

0.36

 torre

0.35

Rosen

0.35

fl

0.35

POSITIVE LOGITS

 issuing

0.51

DICTION

0.44

鬨

0.43

뿌

0.41

 fortification

0.41

 ברי

0.41

 niezbęd

0.41

 wetter

0.41

avkh

0.41

ದ್ಧ

0.40

Activations Density 0.001%