INDEX

Explanations

review policy

The neuron fires on section or field headings (e.g. “Top features:”, “Product Details”, “Review Policy”), i.e. metadata and list‐title labels.

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

and

-2.08

by

-1.85

as

-1.77

or

-1.48

 følgende

-1.31

 with

-1.30

 showcasing

-1.26

 skolen

-1.25

 וגם

-1.24

Both

-1.23

POSITIVE LOGITS

踯

1.66

 모든

1.43

所以在

1.42

鞫

1.42

</i>

1.40

tè

1.37

黾

1.35

雠

1.34

feb

1.34

fea

1.34

Activations Density 0.000%