INDEX

Explanations

describes conditions or problems

The neuron fires on mid-sentence discourse markers and evaluative qualifiers—adverbs or adjectives like “particularly,” “ideally,” “challenging,” “turns out,” and “evidenced” that signal emphasis or transition in the text.

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

tiness

-1.73

zied

-1.59

 boya

-1.55

raisers

-1.53

 واست

-1.51

 etui

-1.41

 moza

-1.41

lez

-1.40

glades

-1.40

 nhat

-1.38

POSITIVE LOGITS

After

1.82

 After

1.59

Although

1.57

While

1.55

 ikke

1.53

During

1.53

Also

1.52

 with

1.51

 Generally

1.48

 hacer

1.46

Activations Density 0.145%