INDEX

Explanations

make suggestions

The neuron strongly activates on imperative verbs or directive phrases (e.g. “work,” “go,” “take,” “try,” “read,” “judge,” “sort”), i.e. commands.

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

 både

-1.56

"];

-1.51

;">

-1.30

"];

-1.28

<bos>

-1.25

 nemlig

-1.25

'>

-1.25

)";

-1.25

}">

-1.23

 afect

-1.21

POSITIVE LOGITS

1.55

 find

1.20

してしまう

1.16

してしまった

1.15

啦

1.15

has

1.13

ディオ

1.13

لقاء

1.10

常見

1.10

 values

1.10

Activations Density 0.041%