INDEX

Explanations

you / us / me

This neuron activates on direct-address pronouns—especially “you” (and related “us”/“we”)—marking where the text speaks directly to the reader.

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

’

-1.86

↵

-1.84

has

-1.72

</em>

-1.68

 славя

-1.67

 joão

-1.64

 ลาย

-1.61

这么说

-1.59

〇〇

-1.59

ając

-1.58

POSITIVE LOGITS

stood

2.53

ou

2.48

 regard

2.41

standing

2.28

 frecuencia

2.25

 whom

2.13

holds

1.95

 Hence

1.95

 ersten

1.94

1.90

Activations Density 0.012%