INDEX

Explanations

subtracting from the

The neuron fires on words that express spatial relations or orientations (e.g. prepositions like “by,” “at,” “in,” “on” and orientation terms like “front” or “back”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

de

0.65

0.62

 eines

0.60

 ενός

0.57

 unor

0.57

 anderer

0.55

 unui

0.53

gad

0.52

 einem

0.51

POSITIVE LOGITS

 этот

1.07

 этой

0.94

这个

0.91

 this

0.89

 इस

0.88

this

0.88

 này

0.88

 этого

0.88

 эту

0.87

この

0.86

Activations Density 0.810%