INDEX

Explanations

2d

The neuron fires on legal‐case reporter volume identifiers (the “number+letter” parts like “2d” in citations such as “33 S.E.2d”).

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

 ſever

-1.53

 medy

-1.52

diese

-1.46

hawatir

-1.45

复制代码

-1.44

 himmel

-1.43

 ""),

-1.38

 parfü

-1.35

 illustre

-1.34

Interes

-1.34

POSITIVE LOGITS

 always

2.27

 never

2.02

of

1.94

 sometimes

1.94

↵↵

1.85

 they

1.79

 still

1.68

 there

1.63

not

1.58

但也

1.54

Activations Density 0.005%