INDEX

Explanations

hangs up, hangups, Dual reporter

The neuron fires on occurrences of “hang”/“hung” paired with “up” (i.e. variants of “hang up”).

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

 seem

-2.27

ありますが

-2.22

 have

-1.95

Which

-1.88

It

-1.79

 affect

-1.78

 intellect

-1.78

 emphas

-1.77

be

-1.73

还

-1.73

POSITIVE LOGITS

 использования

2.08

},{

2.02

スタム

1.99

 continúas

1.97

 estándares

1.89



1.84

🛇

1.82

 plazos

1.79



1.78

-------

1.77

Activations Density 0.002%