INDEX

Explanations

instead / but / rather

The neuron fires on instruction and advice cues—words in the imperative or directive mood (e.g. “Don’t,” “concentrate,” “if you can,” “force yourself,” 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

 without

0.72

Without

0.70

 ohne

0.64

 luego

0.64

 Without

0.64

これにより

0.63

 including

0.62

 incluyendo

0.62

 recor

0.60

 wobei

0.60

POSITIVE LOGITS

而是

1.33

 Instead

1.31

 বরং

1.31

instead

1.22

むしろ

1.22

 instead

1.20

Instead

1.20

 बल्कि

1.19

nor

1.18

あくまで

1.11

Activations Density 0.770%