INDEX

Explanations

lists, numbers, and formatting

The neuron fires most strongly on sentence boundaries and the very first tokens of sentences (including the period “.” itself), i.e. it’s detecting sentence-boundary or sentence-start markers.

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

 समेत

1.30

 ámbitos

1.24

 LGBTQ

1.23

 metaverse

1.20

 inclusivity

1.16

 รวมถึง

1.16

LGBT

1.13

迎来

1.13

 reimag

1.12

 civiles

1.12

POSITIVE LOGITS

 photographs

1.09

 Using

1.08

 Figure

1.06

 scintillation

1.06

 orifice

1.01

 impedances

0.98

Use

0.97

 voltages

0.96

 diagrams

0.96

Fig

0.95

Activations Density 4.346%