INDEX

Explanations

importance, risk, and consequences

The neuron fires on mentions of “stake” or “stakes,” i.e. words signaling risk or what’s at stake.

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

ار

-1.42

sierung

-1.29

 leiden

-1.28

 xxvi

-1.26

#*

-1.26

 Periode

-1.23

鷙

-1.20

恙

-1.20

 lage

-1.19

 continuar

-1.18

POSITIVE LOGITS

 OSError

1.27

 потенциаль

1.27

írez

1.25

 Eindruck

1.24

 preciosa

1.20

väll

1.20

dată

1.20

性的

1.20

for

1.17

 tearDown

1.17

Activations Density 0.053%