INDEX

Explanations

states that

The neuron detects theorem statements in technical/math text—specifically the “Theorem” heading along with its numbering and punctuation that introduce a formal theorem.

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

侂

-2.00

Terima

-1.75

 với

-1.70

SALUD

-1.66

Hoodie

-1.66

aminen

-1.66

orghini

-1.60

дження

-1.60

ői

-1.59

 للمعارف

-1.59

POSITIVE LOGITS

 that

2.27

2.11

 there

1.85

and

1.77

↵

1.73

1.67

if

1.63

1.60

↵↵

1.55

単

1.52

Activations Density 0.010%