INDEX

Explanations

posted by names, reader writes, before justices

This neuron detects lines that attribute authorship or list named participants (e.g. “posted by…,” “An anonymous reader writes,” “Before…”) by spotting proper‐name metadata headers.

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

 silueta

-1.38

 ministra

-1.34

 falda

-1.27

 endforeach

-1.22

 teatral

-1.21

 galeria

-1.20

 asfal

-1.19

 emblema

-1.16

 parlament

-1.15

 století

-1.13

POSITIVE LOGITS

这两

1.30

每個

1.18

幾乎

1.16

 will

1.15

也會

1.13

It

1.13

這些

1.09

曾經

1.06

entendre

1.05

这两个

1.04

Activations Density 0.001%