INDEX

Explanations

universally

The neuron flags non‐finite verb forms—especially participles ending in “-ed” or “-ing.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ilarang

-1.23

狲

-1.23

 pali

-1.22

 fabrik

-1.22

ané

-1.21

ategorien

-1.21

arische

-1.20

ţie

-1.19

Scénario

-1.18

why

-1.16

POSITIVE LOGITS

 універ

1.27

ly

1.20

 близь

1.18

si

1.13

1.12

 tvá

1.11

on

1.09

適量

1.09

1.08

åt

1.07

Activations Density 1.197%