INDEX

Explanations

location or relationship prepositions

The neuron strongly fires on very high‐frequency “function words” (short closed-class tokens like prepositions, articles, and causal particles) across multiple languages.

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

and

-2.94

how

-1.95

 sierpnia

-1.83

who

-1.79

 Jahrhunderts

-1.76

 mereka

-1.69

 from

-1.65

for

-1.65

 jajaja

-1.63

 września

-1.60

POSITIVE LOGITS

1.77

 делает

1.71

何回

1.63

 reichlich

1.59



1.57

 требо

1.53

 nochmal

1.51

前来

1.49

 deaktiviert

1.49

you

1.48

Activations Density 0.010%