INDEX

Explanations

code structure separators

The neuron fires on programmer‐style identifiers—especially method and class names in code (e.g. things you see after dots or as API calls).

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

 douleur

-1.48

 However

-1.32

ıdır

-1.31

 promesse

-1.28

 село

-1.27

峪

-1.25

we

-1.23

 向け

-1.22

 tahun

-1.21

irdre

-1.20

POSITIVE LOGITS

↵

1.45

这下

1.44

我又

1.38

埝

1.36

—

1.36

卧槽

1.33

我真的

1.32

到现在

1.31

 первых

1.31

 präsenti

1.30

Activations Density 0.008%