INDEX

Explanations

HTML and C# code structure

The neuron strongly activates on HTML/XML tags and related angle-bracketed markup (e.g. <head>, <meta>, etc.).

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

as

-1.48

one

-1.42

 because

-1.41

 within

-1.41

by

-1.39

on

-1.39

 makes

-1.39

 from

-1.38

 that

-1.37

at

-1.36

POSITIVE LOGITS

 EVEN

1.38

1.35

………

1.34

 BEING

1.33

 pivotal

1.33

…….

1.31

 …..

1.31

摀

1.30

 grandiose

1.28

 deliciously

1.27

Activations Density 0.067%