INDEX

Explanations

change state or direction

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

/change

-0.12

ìĿ´ìĸ´

-0.09

als

-0.09

 larg

-0.09

 redund

-0.09

 chats

-0.09

ml

-0.08

POSITIVE LOGITS

over

0.19

 tack

0.16

iable

0.12

cate

0.12

overs

0.12

CodeAt

0.11

ante

0.11

able

0.11

course

0.11

 Hands

0.11

Activations Density 0.029%