INDEX

Explanations

made of

The neuron detects passive‐voice constructions centered on “made” (e.g. phrases like “was made of…” or “have been made to…”).

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

 each

-1.77

one

-1.75

and

-1.51

</em>

-1.50

-1.49

when

-1.38

「

-1.37

Each

-1.36

 every

-1.35

 started

-1.35

POSITIVE LOGITS



1.90

 boletín

1.73

 trône

1.72

以外にも

1.69

 juſ

1.67

 cinquante

1.66

 acteur

1.65

有任何

1.61

 Objetivos

1.58

 quatrième

1.57

Activations Density 0.102%