INDEX

Explanations

making statements or predictions

The neuron fires most strongly on past-tense or past-participle verbs (especially longer “–ed” forms).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

how

-1.45

for

-1.36

 what

-1.25

 where

-1.25

who

-1.23

 как

-1.20

why

-1.19

of

-1.19

are

-1.16

Then

-1.15

POSITIVE LOGITS

 urbanas

1.30

 richiesto

1.30

 clás

1.28

 Públicas

1.27

 educativas

1.24

 тобой

1.23

powiedz

1.23

 mejorar

1.16

 professionisti

1.16

 suggested

1.13

Activations Density 0.124%