INDEX

Explanations

prudish, conservative, uptight

New Auto-Interp

Top Features by Cosine Similarity

Configuration

Prompts (Dashboard)

10,000 prompts, 128 tokens each

Dataset (Dashboard)

lmsys/lmsys-chat-1m

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 selfish

-0.11

 Lazy

-0.10

imde

-0.09

undan

-0.09

icker

-0.09

Ã¡jem

-0.09

unh

-0.09

 æĤ

-0.09

Äįer

-0.09

ego

-0.08

POSITIVE LOGITS

Pur

0.26

pur

0.26

upt

0.25

 boring

0.23

 conservative

0.22

pr

0.20

 moral

0.20

 Moral

0.19

upt

0.18

pur

0.18

Activations Density 0.131%