INDEX

Explanations

always make sure

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 seemingly

-0.12

 somehow

-0.12

 apparently

-0.12

 sadly

-0.11

 magically

-0.11

 suddenly

-0.11

 Apparently

-0.11

chi

-0.11

 evidently

-0.11

 complaining

-0.10

POSITIVE LOGITS

 deliberately

0.22

 intentionally

0.22

 carefully

0.22

 purposely

0.22

 consciously

0.22

 careful

0.19

 always

0.16

 avoided

0.16

 conscious

0.15

 deliberate

0.15

Activations Density 0.285%

No Known Activations