INDEX

Explanations

adapted in, game against, learning and

The neuron tends to fire on content words—i.e. the open‐class nouns and verbs that carry substantive meaning—rather than on function words.

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

 which

-1.67

 things

-1.57

 when

-1.45

 októ

-1.34

Algunas

-1.34

Alguns

-1.33

 która

-1.28

ämään

-1.27

 ")");

-1.27

 decreto

-1.26

POSITIVE LOGITS

and

1.45

1.41

 همچنین

1.30

four

1.30

Both

1.27

to

1.27

Either

1.25

1.24

理由は

1.24

：“……”

1.24

Activations Density 0.174%