INDEX

Explanations

complying with a condition

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 Maar

-0.10

 Laud

-0.08

omba

-0.08

 interp

-0.08

 scrut

-0.07

REQ

-0.07

Vu

-0.07

 errone

-0.07

 multid

-0.07

Aer

-0.07

POSITIVE LOGITS

 comply

0.16

 promise

0.13

 abide

0.13

 complying

0.12

 successfully

0.12

 obey

0.12

 complied

0.12

 convinc

0.12

 agreeing

0.12

 agrees

0.12

Activations Density 0.051%

No Known Activations