INDEX

Explanations

places and events

The neuron strongly activates on named-entity tokens—place names, dates, event titles, workshop/program/tournament/method labels, and other proper-noun tokens.

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

and

-2.52

but

-1.91

etc

-1.60

to

-1.52

 then

-1.52

 while

-1.23

 other

-1.21

ですし

-1.20

is

-1.17

 with

-1.13

POSITIVE LOGITS

為に

1.56

 allerede

1.30

谪

1.24

嗐

1.23

っていない

1.23

Preparación

1.19

Ǵ

1.17

ところに

1.16

何が

1.16

änner

1.16

Activations Density 0.113%