INDEX

Explanations

for achieving or failing

The neuron detects negative/failure constructions, especially negation phrases indicating not doing or failing to do something (e.g. “not having done anything,” “failing to,” “forcing me”).

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

 become

-1.35

 people

-1.32

之为

-1.19

 find

-1.18

 però

-1.15

However

-1.12

ika

-1.10

be

-1.09

さんから

-1.06

 लेकिन

-1.05

POSITIVE LOGITS

during

1.29

primarily

1.20

specifically

1.19

 manquer

1.16

 lenker

1.16

Vidéo

1.16

probably

1.15

 mög

1.14

 certaines

1.14

 sommige

1.13

Activations Density 0.064%