INDEX

Explanations

apologies and personal statements

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

 apologies

-0.22

 Sorry

-0.22

 sorry

-0.22

 apology

-0.19

Sorry

-0.19

 apologized

-0.19

sorry

-0.19

 apolog

-0.19

 apologize

-0.19

 apologise

-0.18

POSITIVE LOGITS

 promise

0.12

 Promise

0.09

promise

0.09

lem

0.09

Ga

0.08

 vulnerable

0.08

 reward

0.08

Lem

0.08

 blame

0.08

sem

0.07

Activations Density 0.063%