INDEX

Explanations

safety and security concerns

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

Orn

-0.11

/theme

-0.11

 hopeless

-0.11

orrh

-0.10

 Orth

-0.10

Fut

-0.10

 infertility

-0.10

-che

-0.10

Rai

-0.10

 pena

-0.10

POSITIVE LOGITS

 safety

0.35

 Safety

0.32

Safety

0.30

 safer

0.27

 unsafe

0.25

 safest

0.24

afety

0.22

 safe

0.21

 Safe

0.21

 security

0.20

Activations Density 0.025%

No Known Activations