INDEX

Explanations

WARRANTIES OR CONDITIONS

The neuron primarily activates on tokens written entirely in uppercase.

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

 Kanto

-0.81

本地

-0.80

 Narrow

-0.77

 Near

-0.71

ulk

-0.69

共

-0.68

んですよ

-0.68

Theory

-0.68

 Autop

-0.66

TabbedPane

-0.65

POSITIVE LOGITS

kozás

0.70

strftime

0.68

lexic

0.67

Milano

0.67

 misma

0.66

零

0.65

Slo

0.65

ży

0.64

jualan

0.64

elseif

0.63

Activations Density 0.067%