INDEX

Explanations

cannot load file

The neuron fires on the numeric tokens (and their surrounding colons) that mark line numbers, error codes, or diagnostic positions in compiler/runtime error messages.

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

</h2>

-1.42

 może

-1.40

 ベース

-1.30

 時計

-1.29

CLUSIVE

-1.27

 illustratie

-1.26

)}\

-1.25

 sombrero

-1.21

nda

-1.21

ころ

-1.21

POSITIVE LOGITS

are

1.59

";

1.51

魔力

1.48

 actualmente

1.40

Lähde

1.38

 chief

1.34

 since

1.34

1.32

克服

1.32

Activations Density 0.073%