INDEX

Explanations

issues, options, aspects

The neuron strongly activates on abstract, content-bearing nouns that label key features, metrics, or components (e.g. “trust factor,” “load time,” “important item,” “bike share component”).

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

 संस्करण

0.87

 —,

0.82

𒅌

0.80

iprofloxacin

0.78

 ダンロップ

0.77

 ซึ่ง

0.73

といった

0.73

<0xD5>

0.73

 による

0.73

グ

0.72

POSITIVE LOGITS

0.86

 अत्य

0.83

ление

0.82

кування

0.78

 unending

0.77

uot

0.75

pada

0.74

 sustenance

0.74

ું

0.73

Activations Density 1.128%