INDEX

Explanations

contrasting opinions and beliefs

The neuron responds to words signaling someone’s opinion or claim—terms like “others,” “thought,” “believe,” “claim,” “opinion,” etc., that mark alternative viewpoints.

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

 stratég

-1.52

 ongeveer

-1.38

 również

-1.37

 menek

-1.30

Aun

-1.22

 waarschijnlijk

-1.20

 parfüm

-1.19

失礼

-1.16

 segítség

-1.14

 során

-1.13

POSITIVE LOGITS

 themselves

1.47

 there

1.45

如

1.30

 knows

1.20

 always

1.16

 thinks

1.15

 prefer

1.14

 wants

1.13

 believes

1.13

 like

1.11

Activations Density 0.053%