INDEX

Explanations

can

The neuron strongly activates on the head noun of common multi-word noun compounds or technical terms (e.g. “minimum wage,” “neural networks,” “brand tracking,” “echo chamber,” “political thrillers”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 начали

0.51

ettiin

0.46

 الثالث

0.44

 объяви

0.43

 estavam

0.43

陆续

0.42

 estava

0.41

 الشعب

0.41

 вышел

0.40

이었

0.40

POSITIVE LOGITS

 dapat

0.59

 può

0.59

 pueden

0.56

 bisa

0.56

 possono

0.54

 lahko

0.54

 posso

0.54

 axiomatic

0.54

 galima

0.53

can

0.53

Activations Density 0.000%