INDEX

Explanations

with or without modification

The neuron fires on the key “alternative” term in binary/conditional constructions (e.g. the word after “or without …” or “may or may not … be …”), i.e. it picks out the contrasted item in an “or” clause.

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

and

-2.38

</sup>

-1.69

 pouze

-1.68

 rather

-1.55

 instead

-1.52

 тільки

-1.48

-1.41

 seulement

-1.39

Instead

-1.39

 hlavně

-1.38

POSITIVE LOGITS

 related

1.75

 such

1.57

we

1.54

 those

1.52

 nuove

1.48

of

1.47

1.39

презента

1.36

каль

1.35

 chiare

1.31

Activations Density 0.103%