INDEX

Explanations

The main thing this neuron does is spot numeric or statistical tokens—counts, measurements, percentages, and other quantitative data.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 airfield

0.73

 airlift

0.69

工艺

0.69

 granite

0.67

 Stiffness

0.67

 transonic

0.66

 carbonic

0.65

 تجه

0.65

 strapping

0.65

 trusses

0.64

POSITIVE LOGITS

 Twitter

2.14

Twitter

2.05

 Facebook

1.93

 Tumblr

1.93

Facebook

1.89

 hashtags

1.88

 Snapchat

1.80

 Weibo

1.78

 twitter

1.76

 Reddit

1.73

Activations Density 1.615%

No Known Activations