INDEX

Explanations

sentence endings and conjunctions

This neuron detects strong clause or sentence boundaries—specifically punctuation like periods and dashes (including em-dashes and hyphens) and the tokens immediately following them.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 dupla

-1.75

 enorme

-1.68

᪥

-1.58

࿚

-1.57

Präsident

-1.57

 eles

-1.57

cinoma

-1.54

푆

-1.52

 experimenta

-1.49

 сьогодні

-1.49

POSITIVE LOGITS

The

2.34

and

1.97

or

1.85

1.70

In

1.59

It

1.59

 This

1.57

This

1.50

But

1.47

These

1.43

Activations Density 0.147%