INDEX

Explanations

can do or can be

The neuron strongly fires on occurrences of the modal “can” (especially in “can do”-type ability/possibility phrases).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits



-1.80

桖

-1.66

 antal

-1.60

 blessures

-1.53

 sord

-1.52

✴

-1.50

-1.48



-1.47

饃

-1.47

τητα

-1.47

POSITIVE LOGITS

 that

1.59

 from

1.52

1.46

for

1.43

 both

1.42

on

1.41

でしたか

1.40

 more

1.38

was

1.38

 способствует

1.36

Activations Density 0.011%