INDEX

Explanations

exit or return

This neuron detects instructional or how-to language, i.e. second-person directives and tutorial steps.

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

 کلی

0.61

ITA

0.58

 udało

0.58

łac

0.56

}}}-\

0.56

었다

0.54

 қар

0.54

 ಬ್ಯಾ

0.54

արգ

0.53

↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.53

POSITIVE LOGITS

Skip

0.78

 skip

0.74

 Skip

0.68

skip

0.64

我去

0.64

 skipping

0.56

 loaf

0.55

 SKIP

0.55

 skips

0.54

断

0.54

Activations Density 0.002%