INDEX

Explanations

pronouns followed by past actions

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

 auch

-0.11

 Ð¢Ð°ÐºÐ¶Ðµ

-0.11

alse

-0.11

 also

-0.10

åį´

-0.10

also

-0.10

 cÅ©ng

-0.09

urum

-0.09

ught

-0.09

 anche

-0.09

POSITIVE LOGITS

 totiÅ¾

0.13

inz

0.11

oretical

0.11

aven

0.11

oret

0.10

 Ð²ÐµÐ´ÑĮ

0.10

 zwar

0.10

ary

0.09

 vá»ĳn

0.09

it

0.09

Activations Density 0.057%