INDEX

Explanations

word or phrase following certain nouns

The neuron flags standalone, content-bearing nouns or technical keywords across different contexts (e.g. “Theory,” “reef,” “document,” “observable,” “batches,” “stamp,” “number,” “bridge”).

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

ppling

-1.85

who

-1.55

at

-1.42

 those

-1.34

 vérité

-1.31

They

-1.30

 spou

-1.27

 première

-1.27

asting

-1.26

 الذين

-1.26

POSITIVE LOGITS

 niewiel

1.66

 atividade

1.52

 großem

1.47

}+

1.45

 colchón

1.45

 heuti

1.43

 blisko

1.41

笆

1.41

 tartalmaz

1.40

 najwięks

1.39

Activations Density 0.203%