INDEX

Explanations

caring for

This neuron detects occurrences of the verb “care” (including “caring”) typically followed by “for.”

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

-1.84

鹬

-1.82

）；

-1.77

旆

-1.75

}),

-1.74

ならば

-1.72

 atraer

-1.68

})}

-1.68

ቼ

-1.66

牁

-1.65

POSITIVE LOGITS

 hinder

2.00

 famosas

1.83

也可

1.77

蓊

1.75

it

1.74

ebenarnya

1.74

 dangereux

1.71

1.70

 هذا

1.66

 bunu

1.64

Activations Density 0.006%