INDEX

Explanations

language proficiency and fluency

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

bew

-0.10

 timeless

-0.08

igest

-0.08

 repeat

-0.08

educ

-0.08

SENS

-0.08

 mismatch

-0.08

bip

-0.08

 indirectly

-0.08

 dwell

-0.07

POSITIVE LOGITS

 proficiency

0.24

flu

0.24

 fluent

0.21

Flu

0.20

 rusty

0.18

 proficient

0.18

 convers

0.17

 Fluent

0.17

flu

0.17

 vocabulary

0.17

Activations Density 0.401%

No Known Activations