INDEX

Explanations

first names and last names

The neuron strongly activates on multi‐token proper names—especially personal first (and sometimes last) names—i.e. capitalized name chunks.

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

 their

-1.41

 still

-1.38

-1.33

 española

-1.26

 then

-1.24

as

-1.23

 began

-1.19

</h4>

-1.19

an

-1.15

Ar

-1.14

POSITIVE LOGITS

 upwards

1.30

 daardoor

1.22

ほんと

1.20

 efek

1.20

 Dumnezeu

1.20

 atrak

1.19

larda

1.18

between

1.16

 avena

1.16

興味

1.15

Activations Density 0.097%