INDEX

Explanations

The neuron fires on emphatic review language—especially exclamation points, sentence-ending punctuation, and strongly positive words (e.g. “excellent,” “lots,” “great”).

end of sentence

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

 this

-1.45

möglichen

-1.31

已经

-1.26

 what

-1.25

已

-1.18

 нажмите

-1.13

in

-1.11

or

-1.07

 грудня

-1.05

now

-1.05

POSITIVE LOGITS

was

1.51

 ficou

1.45

 were

1.34

 وكان

1.31

 loved

1.22

Autores

1.22

 necesitar

1.18

ꦢ

1.18

 alugar

1.18

 overall

1.16

Activations Density 0.003%