INDEX

Explanations

as / cómo / when

The neuron fires on content words that are verbs (in various languages)—i.e. it flags verb forms wherever they appear.

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

 concorda

-1.59

 emisiones

-1.32

 adoro

-1.30

 ftate

-1.30

 favorece

-1.27

 provar

-1.27

 vengo

-1.25

 вроде

-1.23

ishima

-1.19

 crescer

-1.19

POSITIVE LOGITS

as

9.94

作为

5.13

 sebagai

5.13

 jako

4.66

 ως

4.28

 как

4.25

作為

4.22

 качестве

4.06

 як

3.66

 به‌عنوان

3.44

Activations Density 0.014%