INDEX

Explanations

likely indicates probability

The neuron flags words and short phrases that serve as discourse connectives or logical/relational markers (e.g. “part of,” “because,” “more,” “related,” “involved,” “still,” etc.).

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

</h2>

-1.83

꜆

-1.70

 تعدى

-1.67

urantes

-1.60

 alimen

-1.55

consola

-1.55

 apparaissent

-1.54

 drape

-1.50

咙

-1.50

 occurrences

-1.48

POSITIVE LOGITS

 antworte

1.83

1.80

1.72

ي

1.63

 vienas

1.62

はお

1.60

1.59

 changed

1.56

一只

1.55

 Schluß

1.55

Activations Density 0.015%