INDEX

Explanations

The neuron fires strongly on standalone headings or topic-sentence lines—i.e. titles, subheadings, or isolated question/statement prompts in the text.

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

 secretions

0.95

 piston

0.83

 torsion

0.81

 semblance

0.81

 по

0.78

 doped

0.78

 necrotic

0.78

 simulates

0.77

 projectiles

0.77

 velocities

0.76

POSITIVE LOGITS

↵↵

1.38

But

1.30

 However

1.30

 Nhưng

1.30

 Although

1.29

 Additionally

1.29

 While

1.23

 Because

1.23

 Furthermore

1.21

But

1.20

Activations Density 4.050%