INDEX

Explanations

returning to a state

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

 Backend

-0.11

rita

-0.10

Backing

-0.09

Repeated

-0.09

Backend

-0.09

.persist

-0.09

ÎºÎ¿

-0.09

lea

-0.09

ãģķãĤī

-0.08

Backdrop

-0.08

POSITIVE LOGITS

 again

0.77

again

0.61

 Again

0.56

Again

0.53

 wieder

0.53

 ëĭ¤ìĭľ

0.49

 láº¡i

0.47

 ÑģÐ½Ð¾Ð²Ð°

0.47

 novamente

0.46

 weer

0.45

Activations Density 0.132%