INDEX

Explanations

Hibernate mapping

The neuron responds strongly to tokens containing digits (including standalone numbers, hex codes in percent-escapes, version numbers, IDs, etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

on

-1.77

in

-1.70

 having

-1.66

all

-1.66

 every

-1.60

 when

-1.55

or

-1.52

 meticulously

-1.49

 encompasses

-1.48

 from

-1.45

POSITIVE LOGITS

 unending

1.63

 gigantic

1.57

 wondrous

1.56

 idiotic

1.52

1.50

 intriguing

1.44

 diversos

1.43

 dessa

1.40

 myriad

1.39

ch

1.38

Activations Density 0.001%