INDEX

Explanations

slightly

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

ãĤĤãģ£ãģ¨

-0.09

///<

-0.09

low

-0.09

hoe

-0.08

xes

-0.08

REW

-0.08

 Sant

-0.08

.mj

-0.08

 Sind

-0.08

Neal

-0.08

POSITIVE LOGITS

 slightly

0.70

 slight

0.66

ç¨į

0.38

 slightest

0.35

lightly

0.34

çķ¥

0.32

 margin

0.31

sl

0.29

 leicht

0.29

 Ð½ÐµÐ¼Ð½Ð¾Ð³Ð¾

0.29

Activations Density 0.160%