INDEX

Explanations

The neuron fires strongly on capitalized named entities (brand names, product names, place names, etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 oWord

0.76

 História

0.76

امية

0.75

ะ

0.73

AcOH

0.73

不會

0.71

Спољашње

0.70

 Địa

0.70

)，

0.70

 Royaume

0.69

POSITIVE LOGITS

has

0.82

0.77

rutt

0.72

 angg

0.71

 Technically

0.69

 സര്‍ക്കാര്‍

0.69

th

0.67

<b>

0.67

tiered

0.67

 घेर

0.67

Activations Density 0.001%