INDEX

Explanations

lists of features

The main thing this neuron does is pick out rare or highly informative content words (i.e. keywords) rather than common function words.

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

 nelle

0.90

 こんな

0.88

 flamboyant

0.86

 epiphany

0.82

 torso

0.82

 handlebar

0.82

 concealer

0.82

 никакой

0.82

 meringue

0.81

 conceit

0.81

POSITIVE LOGITS

نن

0.77

.$

0.74

Although

0.71

یا

0.71

🌉

0.71

вета

0.69

require

0.69

tf

0.67

نية

0.67

ن

0.66

Activations Density 0.000%