INDEX

Explanations

alphanumeric identifiers

The neuron strongly activates on hexadecimal‐digit tokens (letters A–F and numerals) within long ID or hash strings.

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

for

-1.96

on

-1.72

 even

-1.46

-1.41

-1.38

All

-1.35

or

-1.34

For

-1.34

-1.33

POSITIVE LOGITS

athered

1.52

Ouverture

1.48

Chaque

1.46

durante

1.45

猻

1.45

鷽

1.41

 RELIG

1.39

Probl

1.38

 Cruelty

1.37

ジナル

1.36

Activations Density 0.001%