INDEX

Explanations

home or household phrases

This neuron activates most strongly on uncommon or low‐frequency “content” tokens—that is, rare or unusually informative words.

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

↵↵↵

-1.98

↵↵

-1.91

↵↵↵↵

-1.88

↵↵↵↵↵

-1.81

。

-1.67

↵↵↵↵↵↵

-1.58

↵↵↵↵↵↵↵

-1.53

 más

-1.52

та

-1.50

↵↵↵↵↵↵↵↵↵

-1.48

POSITIVE LOGITS

 those

1.80

jacent

1.45

㈱

1.43

 totalité

1.42

Eſ

1.38

日を

1.37

 ανθρώ

1.36

へん

1.35

 považ

1.34

 bricola

1.34

Activations Density 0.067%