INDEX

Explanations

more + adverb

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

assic

-0.10

byn

-0.10

safe

-0.10

mer

-0.10

 swift

-0.09

byss

-0.09

 Claud

-0.09

icable

-0.09

 assez

-0.08

ennent

-0.08

POSITIVE LOGITS

Ð¾Ð¹Ð½Ð¾

0.15

 easily

0.15

-effect

0.13

ÑĢÐ°ÑĤÐ½Ð¾

0.13

 effectively

0.13

 quickly

0.12

 than

0.12

izontally

0.12

ÑĸÐ¹Ð½Ð¾

0.11

ursively

0.11

Activations Density 0.039%