language and instruction

response demonstrates self-awareness

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

0.63

were

0.50

0.48

Then

0.48

0.47

and

0.47

well

0.46

POSITIVE LOGITS

ativa

0.59

ivă

0.49

 partur

0.48

IDUAL

0.48

avacanam

0.47

 convocatoria

0.47

avacan

0.46

 demostrado

0.46

 autorizado

0.46

ጨም

0.45

Activations Density 0.001%