INDEX

Explanations

instructions and warnings

The neuron detects instructional or advisory phrasing—markers like “how to,” “should,” “needs,” “First,” etc., that signal step-by-step guidance or recommendations.

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

typical

-0.81

Forex

-0.81

###############

-0.79

######

-0.79

 ВО

-0.78

нав

-0.78

 '-':

-0.77

Γ

-0.77

 benne

-0.77

句话

-0.76

POSITIVE LOGITS

 remember

1.31

注意

1.26

 must

1.23

 trebuie

1.23

pay

1.21

 attention

1.13

 注意

1.13

 beware

1.12

Remember

1.10

 grasp

1.05

Activations Density 0.007%