INDEX

Explanations

next step

The neuron fires on pagination and navigation cues—especially “Next,” “>,” and numeric page‐ or item‐indicator tokens.

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

hradun

-1.25

 dunia

-1.19

 keineswegs

-1.18

 tuer

-1.16

 akar

-1.15

---*/

-1.14

QMessageBox

-1.13

'],

-1.13

 británica

-1.11

躂

-1.11

POSITIVE LOGITS

how

1.59

 Forder

1.59

 howling

1.29

 ощу

1.25

这里

1.24

]$.

1.23

 его

1.18

他

1.16

 jungles

1.16

都是

1.16

Activations Density 0.007%