INDEX

Explanations

Thanks

The neuron fires on the “A:” answer header and on the trailing “Thanks.” signature (i.e. markers that denote the start of an answer or the end‐of‐post sign‐off).

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

by

-1.36

igraphy

-1.34

다는

-1.20

 CERTAIN

-1.19

 pensieri

-1.19

 messaggi

-1.18

heiros

-1.18

 Otras

-1.16

of

-1.16

 densité

-1.13

POSITIVE LOGITS

who

1.63

 considerar

1.37

任何

1.25

鉺

1.23

 mirar

1.23

Their

1.20

 республики

1.20

༑

1.19

ectores

1.18

schwar

1.18

Activations Density 0.309%