INDEX

Explanations

recommendations or requirements

This neuron detects phrases expressing obligation, recommendation, or necessary actions (e.g. “needs to be given,” “should,” “get it fixed”).

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

 거의

0.84

 několika

0.76

踩

0.76

 několik

0.75

 अगदी

0.73

几乎

0.73

 unknowingly

0.72

 yaklaşık

0.71

 meestal

0.71

샀

0.71

POSITIVE LOGITS

 deberían

1.38

 urgently

1.37

 debería

1.36

 powinno

1.30

 powin

1.27

 should

1.23

 dovrebbe

1.23

 devraient

1.21

should

1.20

 dringend

1.20

Activations Density 1.191%