INDEX

Explanations

punishment and discipline

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

 suicidal

-0.11

 unprotected

-0.11

 scanned

-0.11

 explor

-0.11

 volunteering

-0.11

 greedy

-0.10

 Rings

-0.10

 solic

-0.10

 acquisitions

-0.10

Gre

-0.10

POSITIVE LOGITS

 punish

0.37

 punishments

0.36

 punishment

0.36

 punitive

0.36

 punishing

0.34

 discipl

0.33

 disciplinary

0.30

 discipline

0.28

 penal

0.27

Pun

0.27

Activations Density 0.225%