INDEX

Explanations

a/an time duration

The neuron specifically detects the indefinite article “an.”

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

我不知道

-1.97

歃

-1.87

 便利

-1.81

Frauen

-1.80

 ordf

-1.78

为啥

-1.77

撄

-1.76



-1.72

 ofta

-1.69

 menyen

-1.69

POSITIVE LOGITS

 that

2.22

 only

2.06

2.05

and

1.91

There

1.81

 also

1.71

1.65

 last

1.64

Only

1.62

because

1.54

Activations Density 0.015%