INDEX

Explanations

received from

The neuron fires on abstract promotional or persuasive nouns (e.g. “applications,” “encouragement,” “support,” “evidence,” “investment,” “help,” “quotes,” etc.) that often appear in calls to action or marketing appeals.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

hstack

0.54

 অত

0.54

"$

0.50

圓

0.49

所に

0.48

colorPrimary

0.48

aisy

0.48

 मामूली

0.48

 using

0.47

嚆

0.47

POSITIVE LOGITS

from

2.67

 from

2.59

 từ

2.44

 FROM

2.36

จาก

2.33

 från

2.31

 dari

2.09

FROM

2.08

 फ्रॉम

2.00

 από

1.97

Activations Density 0.637%