INDEX

Explanations

all but one

The neuron detects numeric exception counts in phrases of the form “all but <N>,” i.e. the number of items excluded.

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

 двох

-1.53

řízen

-1.48

extrémité

-1.46

 кілька

-1.45

Heraus

-1.44

ご覧

-1.43

盉

-1.42

Beberapa

-1.41

诅

-1.41

 コート

-1.40

POSITIVE LOGITS

of

2.92

are

1.98

1.72

1.59

who

1.47

 would

1.47

 that

1.45

1.37

The

1.33

”.

1.32

Activations Density 0.020%