INDEX

Explanations

legalistic terms, technical words

The neuron fires on uncommon or domain‐specific content words (i.e. rare technical, legal, scientific, or programming terms) rather than on high‐frequency function words.

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

 Públicas

-1.46

龊

-1.25

をしている

-1.22

ריאה

-1.21

 चीज़

-1.20

 prega

-1.20

 intende

-1.17

 breathtaking

-1.17

大丈夫です

-1.15

vestimento

-1.13

POSITIVE LOGITS

),

1.21

\%,

1.20

}},

1.16

靄

1.06

if

1.03

but

1.02

然後

1.02

°;

1.02

郭

1.02

="",

1.01

Activations Density 0.373%