INDEX

Explanations

`const` or `(` followed by specific terms

The neuron fires strongly on content‐bearing words that are relatively rare or technical in the text (i.e. low‐frequency, non‐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

י

0.99

abhavam

0.90

कल

0.89

৬

0.88

্ু

0.86

מי

0.84

बई

0.83

wani

0.81

akuan

0.81

所以在

0.80

POSITIVE LOGITS

ђу

0.79

shells

0.73

 oblig

0.71

Beware

0.71

 मांगी

0.70

flows

0.70

 Disposition

0.68

 анализа

0.67

 disintegr

0.67

 Δια

0.67

Activations Density 0.000%