INDEX

Explanations

example of a

The neuron is looking for structural or “label” tokens in technical/instructional contexts—things like list‐item numbers (Step 1, Scene 1), code keywords or identifiers (local, require), and entity/table names in a schema.

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

了自己的

-1.56

 only

-1.56

了一個

-1.51

and

-1.43

The

-1.42

了些

-1.40

了自己

-1.38

-1.31

 after

-1.28

 Even

-1.27

POSITIVE LOGITS

牐

1.80

.《

1.61

܇

1.59

fillment

1.55

嵓

1.52

 murale

1.52

 Atención

1.52

まずは

1.52

aneity

1.50

また

1.44

Activations Density 0.022%