INDEX

Explanations

citation markers like [1] or [@B1]

The neuron reliably fires on bracketed citation or reference markers (e.g. “[52.5]”), indicating it’s detecting in-text citation notations.

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

its

-2.48

 them

-2.34

 verantwort

-1.75

 Verantwort

-1.69

 italianos

-1.61

 verkau

-1.51

 Parteien

-1.50

 của

-1.50

the

-1.49

 trouvez

-1.48

POSITIVE LOGITS

 أيضاً

1.73

 също

1.69

ensitive

1.58

这里

1.48

vì

1.46

ilisation

1.44

อีก

1.41

atical

1.41

Also

1.38

llllll

1.33

Activations Density 0.002%