INDEX

Explanations

LevelDB Authorsyap database values

The neuron strongly activates on capitalized identifiers and names—e.g. library/module names, acronyms, or other multi‐letter uppercase 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

 potrà

-1.49

ご注意

-1.30

 avevano

-1.29

洐

-1.28

eschön

-1.24

 gleiche

-1.24

糗

-1.24

 blauw

-1.23

 다른

-1.23

 fikk

-1.22

POSITIVE LOGITS

」，

1.21

 contrário

1.18

 tâm

1.17

1.14

部は

1.13

⿻

1.12

 لكن

1.11

you

1.10

tiglie

1.09

 ocup

1.08

Activations Density 0.031%