INDEX

Explanations

FTC COMMISSION vex matrix xml addData

The neuron fires on named entities and technical jargon tokens—things like organization names, acronyms, file‐format labels, and domain-specific keywords.

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

طيع

-1.43



-1.38

 său

-1.34

😮

-1.28

and

-1.28

saurus

-1.27

but

-1.24

its

-1.23

 estup

-1.20

 densidad

-1.20

POSITIVE LOGITS

嵛

1.52

uſe

1.49

 некоторы

1.42

蒾

1.40

</strong>

1.39

訫

1.37

͔

1.34

美味しかった

1.33

どの

1.30

 மற்ற

1.30

Activations Density 0.002%