INDEX

Explanations

denying specific information

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

 Fence

-0.10

TBD

-0.10

 rather

-0.09

 Reyes

-0.09

 instead

-0.09

 Floyd

-0.09

nech

-0.08

ga

-0.08

udeau

-0.08

POSITIVE LOGITS

nor

0.36

nor

0.24

Nor

0.20

 unless

0.19

 exact

0.16

 neither

0.16

Nor

0.15

NOR

0.15

exact

0.15

unless

0.14

Activations Density 0.177%