INDEX

Explanations

how far will it go

The neuron detects interrogative structures by firing strongly on question words (e.g. what, when, how, do, can) and associated auxiliary or “to” tokens that signal a question.

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

 panneaux

-1.31

 thing

-1.30

 båda

-1.29

⁶

-1.25

 meubles

-1.23

 ДО

-1.20

ným

-1.20

 escaleras

-1.18

 carreaux

-1.17

 haza

-1.16

POSITIVE LOGITS

of

2.25

 Situationen

1.30

 some

1.27

 inflicting

1.26

 verbringen

1.23

 Empfehlungen

1.23

 have

1.22

 anklicken

1.20

 poszczegól

1.17

 rumahnya

1.16

Activations Density 0.017%