INDEX

Explanations

given, original, input

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

ersist

-0.09

 rever

-0.09

 Blond

-0.09

abcdefghijklmnop

-0.08

/respond

-0.08

ervas

-0.08

isContained

-0.08

 èĸ

-0.08

imli

-0.08

imei

-0.08

POSITIVE LOGITS

 original

0.12

 given

0.11

given

0.10

 input

0.09

 gá»ĳc

0.09

original

0.09

.original

0.09

 Input

0.09

 Given

0.09

Activations Density 0.163%