INDEX

Explanations

calibrated

This neuron fires on subword tokens that include the invisible “soft hyphen” marker.

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

 Что

0.38

Killing

0.32

Elephant

0.32



0.32

Desert

0.31

Calif

0.31

Armstrong

0.31

Funeral

0.31

Big

0.31

SEPT

0.31

POSITIVE LOGITS

 hota

0.28

oretically

0.27

hry

0.26

 hoti

0.25

 discourses

0.25

 enveloped

0.25

 linear

0.24

್ಯಾ

0.24

 noten

0.24

 emotive

0.24

Activations Density 0.009%