INDEX

Explanations

pixel dimensions and sizes

The neuron fires on numeric tokens (literal numbers) in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

bemos

-1.12

Võ

-1.08

nemy

-1.07

thaf

-1.04

 Clé

-1.03

nior

-1.03

 الأكثر

-1.02

eddings

-0.99

 Kalau

-0.97

cleos

-0.97

POSITIVE LOGITS

 ebenfalls

1.22

↵↵↵↵

1.13

↵↵↵↵↵↵↵↵↵↵↵

1.11

↵↵↵↵↵↵↵↵↵↵

1.08

 verschiedene

1.07

↵↵↵↵↵↵↵↵↵↵↵↵↵

1.06

 also

1.02

 offrire

1.02

↵↵↵↵↵↵↵↵

1.00

↵↵↵↵↵↵↵↵↵↵↵↵

1.00

Activations Density 0.033%