INDEX

Explanations

betrayal and trust

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

 awkward

-0.13

 veto

-0.11

STYLE

-0.11

 summons

-0.10

 nons

-0.10

 weird

-0.10

bins

-0.10

ecl

-0.10

ror

-0.10

 embarrassing

-0.10

POSITIVE LOGITS

 betrayal

0.22

 trustworthy

0.20

 betrayed

0.20

 trusting

0.20

 betray

0.19

 trust

0.18

 Trust

0.18

 betr

0.18

 trusted

0.18

trust

0.17

Activations Density 0.096%