INDEX

Explanations

The neuron activates on numbered list markers and section-heading tokens (e.g. “1)”, “2)”, “Segment Your List –”, “Promote Events –”), i.e. enumeration and heading labels.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

and

0.31

but

0.30

 pero

0.29

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.27

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.27

 কিন্তু

0.27

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.26

↵↵↵↵↵↵

0.25

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.25

]]

0.25

POSITIVE LOGITS

 utilizzare

0.29

 використання

0.28

 provide

0.27

 evitare

0.27

 ensures

0.27

 penggunaan

0.27

 ensure

0.26

避免

0.26

 éviter

0.26

 avoid

0.26

Activations Density 0.646%