INDEX

Explanations

descriptive clauses and conditional statements

The neuron is highlighting semantically important “content” tokens—key nouns, verbs, numbers, etc.—that carry the main informational load of the text rather than common function words.

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

-1.50

 empezaron

-1.40

 gòn

-1.27

 poté

-1.27

 obuv

-1.25

URLEncoder

-1.23

$,

-1.20

 kateg

-1.19

URLException

-1.18

 Befes

-1.18

POSITIVE LOGITS

 comunica

1.41

 maniere

1.32

to

1.27

 ultime

1.26

 festge

1.24

 comanda

1.22

 diversa

1.22

 فره

1.21

 verstär

1.20

本书

1.20

Activations Density 0.921%