INDEX

Explanations

The neuron is tuned to pick up on identity or role labels—single-word descriptors like “gay,” “activist,” “director,” etc.

date announcements

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 daarna

-1.28

 publiek

-1.13

 posibilidades

-1.08

ikala

-1.05

が無い

-1.03

不好意思

-1.02

 می‌تواند

-1.02

 pueden

-1.00

 אנחנו

-1.00

POSITIVE LOGITS

 видеть

1.20

++;

1.20

Når

1.16

1.13

.+\

1.09

Ikke

1.08

日時

1.05

addTo

1.03

 coloro

1.02

 不可

1.02

Activations Density 0.001%