INDEX

Explanations

any would think

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

 totally

-0.11

 hearts

-0.10

ActivityResult

-0.10

 completely

-0.09

ilan

-0.09

/respond

-0.09

 compromising

-0.08

abin

-0.08

Lud

-0.08

 SUPER

-0.08

POSITIVE LOGITS

 even

0.19

 anything

0.17

even

0.15

ANY

0.14

 existence

0.14

 anywhere

0.13

 nawet

0.12

anything

0.12

 Ð´Ð°Ð¶Ðµ

0.12

 Anything

0.12

Activations Density 0.060%