INDEX

Explanations

demons, disaster, Son, answer

The neuron fires on words that introduce warnings or directives—imperative verbs and modal constructions used to caution or instruct (e.g. “Why…should,” “Watch out for,” “will appear”).

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

美国的

-1.59

jących

-1.34

</td>

-1.33

 ALWAYS

-1.27

 ilang

-1.23

然后再

-1.21

PUESTA

-1.21

inaires

-1.20

 должна

-1.20

}:

-1.20

POSITIVE LOGITS

鲩

1.33

 увели

1.16

 aside

1.14

iele

1.09

 apart

1.07

 इन

1.05

 estos

1.05

 hydraulic

1.05

1.03

ৢ

1.03

Activations Density 0.052%