INDEX

Explanations

end of turn, start of response

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

 newcom

-0.08

[,]

-0.08

{@

-0.08

ìĪľ

-0.08

[js

-0.08

(#)

-0.08

inke

-0.08

 UIAlert

-0.08

elden

-0.08

WithValue

-0.08

POSITIVE LOGITS

ousel

0.09

lian

0.09

adan

0.09

lich

0.08

tou

0.08

hold

0.08

ekl

0.08

Tou

0.08

uto

0.08

 bgcolor

0.08

Activations Density 0.167%