INDEX

Explanations

The neuron predominantly activates on tokens that convey numerical or quantitative information—numbers, percentages, measurements, and related quantitative terms.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ри

1.45

ü

1.27

ishment

1.20

 الى

1.13

 Bản

1.12

izienz

1.09

 marvelous

1.09

ní

1.08

 jiwa

1.08

 prioritizing

1.07

POSITIVE LOGITS

 Pals

1.50

ोत्तर

1.42

PHE

1.41

ALSO

1.37

TTY

1.37

वर्ती

1.35

สภาพ

1.30

DID

1.28

 Pleas

1.27

eches

1.23

Activations Density 6.837%