INDEX

Explanations

mass graves, exceptions, activity, distributed, size

The neuron fires most strongly on unusually informative or content-heavy words (i.e. rare or high-information tokens).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 segíts

-1.38

 where

-1.37

xes

-1.36

 Obtén

-1.33

 zarówno

-1.32

 теперь

-1.30

 since

-1.27

 Теперь

-1.26

افة

-1.25

 szinte

-1.25

POSITIVE LOGITS

1.35

lossene

1.35

usar

1.27

 immedi

1.25

1.23

',

1.22

 pincel

1.20

↵↵↵

1.19

 αλλά

1.16

 viande

1.16

Activations Density 0.058%