INDEX

Explanations

critical consideration

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

 scorn

-0.09

<quote

-0.09

Ã©gor

-0.09

turnstile

-0.09

overn

-0.09

Æ¯á»

-0.08

belief

-0.08

 prag

-0.08

UFFIX

-0.08

mensagem

-0.08

POSITIVE LOGITS

 critical

0.42

 Critical

0.38

 critically

0.34

Critical

0.34

critical

0.34

 ÐºÑĢÐ¸ÑĤ

0.31

 krit

0.27

 crit

0.26

 critique

0.25

_critical

0.24

Activations Density 0.091%