INDEX

Explanations

The neuron strongly activates on sequences of uppercase letters (acronyms, initialisms, and other all‐caps tokens).

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

 αρκε

0.23

hearted

0.23

 cyclical

0.23

 mundane

0.22

すごく

0.22

唿

0.22

模特

0.22

 handful

0.22

 tidak

0.21

मतौर

0.21

POSITIVE LOGITS

ä

0.27

astră

0.24

세요

0.23

ari

0.23

 Secured

0.23

atter

0.23

ρίου

0.22

arta

0.22

ärke

0.22

äger

0.22

Activations Density 0.625%