INDEX

Explanations

following "the", "a", "construction", "list"

The neuron fires strongly on numeric tokens (especially decimal numbers), so it’s detecting explicit numeric measurements or data values.

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

 Noida

0.86

 Airtel

0.82

 Jharkhand

0.82

 Rohingya

0.81

 Maharashtra

0.80

 lakhs

0.79

 Paytm

0.79

 Telangana

0.79

 Redmi

0.78

 Karnataka

0.77

POSITIVE LOGITS

実験

0.70

عة

0.67

 книга

0.67

ῶν

0.66

 آزمایش

0.63

तिरिक्त

0.63

кова

0.61

 ペット

0.61

نت

0.59

ج

0.59

Activations Density 0.024%