INDEX

Explanations

can do

The main thing this neuron does is pick up on promotional or call-to-action phrases (e.g. “check out,” “read my,” “click here,” etc.).

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

 обеспечи

0.82

 забезпе

0.79

 unoccupied

0.77

 résulte

0.75

 χωρίς

0.75

 обеспечивает

0.74

 exudes

0.73

 ಏನ

0.73

 allegiance

0.71

 tacit

0.71

POSITIVE LOGITS

他の

1.11

 other

0.98

的其他

0.98

 другие

0.97

 OTHER

0.96

其他的

0.94

 अन्य

0.94

other

0.93

 других

0.93

他にも

0.93

Activations Density 0.272%