INDEX

Explanations

consider when or before

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

 belli

-0.09

 beaut

-0.09

assa

-0.09

illis

-0.09

 Graves

-0.08

agg

-0.08

illac

-0.08

ORK

-0.08

 Punch

-0.08

é¢

-0.08

POSITIVE LOGITS

 when

0.29

when

0.23

 before

0.22

khi

0.20

 quando

0.18

 nÃ¤r

0.18

 carefully

0.18

æīįèĥ½

0.17

before

0.17

 ÙĩÙĨÚ¯Ø§Ùħ

0.17

Activations Density 0.061%