INDEX

Explanations

previously unreleased, unpublished, undescribed

The neuron fires on cues of exclusivity or novelty—that is, on words and phrases like “exclusive,” “new,” “unreleased,” “previously,” and “never before.”

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

-1.44

%;

-1.41

 your

-1.27

 AÑO

-1.27

</h2>

-1.27

IS

-1.25

-1.20

-1.19

to

-1.16

POSITIVE LOGITS

ніше

1.25

氏が

1.17

纷

1.11

Lugar

1.10

 dizaines

1.09

 shortness

1.08

izaron

1.08

紛

1.07

 ἐν

1.07

 from

1.06

Activations Density 0.105%