INDEX

Explanations

introspection and awareness

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Beaut

-0.10

Tre

-0.10

 dispens

-0.10

 witness

-0.10

 Witness

-0.09

DITION

-0.09

 Beard

-0.09

Disp

-0.09

 facing

-0.09

 punching

-0.09

POSITIVE LOGITS

 conscious

0.13

 Conscious

0.11

 awareness

0.11

 inspect

0.11

 realization

0.11

Aware

0.11

 Awareness

0.10

 checked

0.10

 consciousness

0.10

 intros

0.10

Activations Density 0.034%