INDEX

Explanations

comment by or posted by

The neuron consistently activates on numeric metadata and comment-header tokens (e.g. comment IDs, scores, ratings, and other numbers) rather than on the main 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

ichtete

-1.47

/>

-1.44

utados

-1.37

ㅣ

-1.35

erializer

-1.34

gradle

-1.31

':'

-1.30

Appellant

-1.30

--

-1.27

POSITIVE LOGITS

the

2.09

it

2.06

are

1.49

of

1.36

 aplique

1.34

 what

1.34

Wady

1.30

でも

1.25

 garante

1.23

 ARISING

1.23

Activations Density 0.013%