INDEX

Explanations

figure, graph, table, data, or code

The neuron fires on words that introduce or refer to accompanying content or examples (e.g. “following,” “given,” “code,” “snippet,” “figure,” “graph,” “table,” “structure,” “sequence”).

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

 reconnaître

-1.55

嫱

-1.45

咴

-1.43

 procéder

-1.40

 comprendre

-1.39

addPreferred

-1.36

 déterminer

-1.34

 بدون

-1.32

 soumettre

-1.32

 ENTIRE

-1.30

POSITIVE LOGITS

for

1.57

 during

1.55

1.52

 such

1.41

one

1.30

 several

1.28

 these

1.27

 some

1.27

by

1.27

During

1.26

Activations Density 0.250%