INDEX

Explanations

listing including

The neuron detects sentence-ending tokens—i.e. it strongly activates on the final punctuation (and immediately adjacent words) at the ends of sentences.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 akku

-1.59

carus

-1.59

aried

-1.53

ebly

-1.47

分から

-1.46

 出演

-1.43

tadiene

-1.39

だったので

-1.38

 semarang

-1.38

haviours

-1.37

POSITIVE LOGITS

of

1.81

but

1.66

 إلا

1.41

When

1.41

<sup>

1.38

can

1.36

ángulo

1.30

 próprias

1.30

 when

1.30

Its

1.28

Activations Density 0.297%