INDEX

Explanations

Name

This neuron is strongly activated by isolated single-character tokens (especially uppercase letters and punctuation marks).

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

 hypersonic

0.63

していますが

0.60

 iconic

0.59

 clás

0.57

 உலகின்

0.57

 বিশ্বের

0.56

 geopol

0.55

 spectacularly

0.55

经典的

0.55

 spectacular

0.55

POSITIVE LOGITS

李

0.82

刘

0.73

 Елена

0.72

 احمد

0.72

 Samantha

0.70

Joshua

0.70

 Joshua

0.70

 Nguyễn

0.69

陈

0.69

陳

0.69

Activations Density 0.393%