INDEX

Explanations

NAME_X followed by verbs or possessives

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

à¹Ħà¸ĭ

-0.10

ã

-0.09

 Separator

-0.09

alles

-0.08

aeper

-0.08

eza

-0.08

uli

-0.08

 Boeh

-0.08

 sscanf

-0.08

kB

-0.08

POSITIVE LOGITS

âĢ²

0.15

âĢ³

0.14

deg

0.10

bis

0.10

 pres

0.10

 Hills

0.09

teenth

0.09

rote

0.09

 Civil

0.09

 aforementioned

0.09

Activations Density 0.057%