INDEX

Explanations

point, Election Day, ground zero

The neuron detects words that signal spatial measurements or levels (e.g. “near,” “half,” “ground,” “sea level,” “walking distance,” etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

of

-2.31

you

-1.72

ОВО

-1.45

 двух

-1.38

 trouvé

-1.37

 наших

-1.35

を選

-1.32

-1.30

 difficiles

-1.30

会被

-1.30

POSITIVE LOGITS

 this

1.89

and

1.81

 when

1.54

REEK

1.39

 erläut

1.38

</em>

1.35

 smallest

1.31

but

1.30

appId

1.30

)”.

1.30

Activations Density 0.132%