INDEX

Explanations

code constructs like `()`

The neuron activates on programming identifiers—tokens that look like function or variable names (e.g. camelCase, PascalCase, or underscore‐separated names).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 以及

0.68

,..

0.64

ामान्य

0.64

 그리고

0.62

 Chances

0.62

 …..

0.62

}:$

0.62

 chứ

0.61

ॉर्डिंग

0.61

‏:‏

0.61

POSITIVE LOGITS

("

1.63

1.59

('

1.56

().

1.55

([

1.47

()[

1.45

("",

1.45

()+"

1.38

())

1.37

("./

1.34

Activations Density 0.022%