INDEX

Explanations

`n` followed by `casecmp` or `Impl`

This neuron detects command‐line option flags (tokens like “-n”, “-ne”, etc.).

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

↵

-1.86

there

-1.85

 fiore

-1.63

 paio

-1.55

 guardi

-1.52

Just

-1.48

 pubb

-1.46

 bambino

-1.43

 árabe

-1.40

 There

-1.39

POSITIVE LOGITS

“,

1.64

be

1.62

🫃

1.53

):

1.44

}:

1.41

)++;

1.41

стройка

1.39

);

1.38

“

1.38

𝕣

1.35

Activations Density 0.004%