INDEX

Explanations

common mistakes and misunderstandings

The main thing this neuron does is detect words expressing omission or things being overlooked (e.g., overlook, miss, fail).

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

 premiered

0.72

했으며

0.71

was

0.68

 היה

0.68

 تمامی

0.68

 تھا

0.68

 честь

0.67

 شهد

0.65

WAS

0.64

HAD

0.64

POSITIVE LOGITS

 mistakenly

1.47

 souvent

1.33

 misconception

1.29

 unknowingly

1.27

 unwittingly

1.27

 unconsciously

1.26

 misunder

1.25

 misconceptions

1.25

误

1.23

 mistaken

1.21

Activations Density 0.871%