INDEX

Explanations

state or condition

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

Styled

-0.11

 Schul

-0.10

 Gins

-0.09

 Styled

-0.09

irit

-0.09

 Ð¸Ð½ÑĦÐ¾ÑĢÐ¼Ð°

-0.09

Č\n

-0.08

undler

-0.08

 precedence

-0.08

POSITIVE LOGITS

 state

0.38

 condition

0.31

 form

0.28

state

0.28

 estado

0.24

çĬ¶æĢģ

0.23

çĬ¶æħĭ

0.21

 ÑģÐ¾ÑģÑĤÐ¾ÑıÐ½Ð¸Ðµ

0.21

 configuration

0.21

 Condition

0.21

Activations Density 0.080%