INDEX

Explanations

Bart names

The neuron fires on name fragments beginning with “Bart,” i.e. proper-name tokens that start with the prefix “Bart.”

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

-2.28

 oscuro

-2.14

-2.08

 répand

-1.98

 😍

-1.98

栊

-1.87

白天

-1.85

 privada

-1.84

 😀

-1.83

但

-1.81

POSITIVE LOGITS

</strong>

2.64

”،

2.42

 unsere

2.33

}$,

2.20

'),

2.16

</th>

2.09

the

2.08

鯧

2.06

 był

2.03

Activations Density 0.001%