INDEX

Explanations

well followed by descriptors

The neuron fires specifically on occurrences of the word “well.”

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

笮

-2.11

ⵌ

-1.93

-1.80

</u>

-1.76

 donner

-1.64

퐹

-1.59

 belki

-1.55

 ricas

-1.55

 baskılı

-1.54

してる

-1.53

POSITIVE LOGITS

方法

1.86

 техні

1.68

because

1.67

</h2>

1.66

्य

1.65

Any

1.64

 незавер

1.63

VIATIONS

1.59

数据

1.59

ència

1.58

Activations Density 0.010%