INDEX

Explanations

that's

The neuron fires on short “wrap-up” or emphasis phrases—things like “so there you have it,” “and that’s …,” or “those are …” that signal a summary or conclusion.

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

as

-1.73

is

-1.62

 bereits

-1.47

 уже

-1.45

He

-1.40

by

-1.38

atypes

-1.36

↵↵

-1.35

 его

-1.34

 already

-1.29

POSITIVE LOGITS

 clôture

1.49

1.30

 レッド

1.27

揹

1.20

 have

1.20

므로

1.17

どれも

1.17

Either

1.16

 récolte

1.16

Activations Density 0.010%