INDEX

Explanations

Khalifa, Rashid, shogun, Al

The neuron signals on subword pieces of proper names and other capitalized named-entity tokens.

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

 wuß

-2.47

 eigenes

-2.22

 mogą

-2.13

 mußten

-2.09

 実例

-1.97

 sahen

-1.97

 ويمكن

-1.95

 zeigten

-1.95

--"

-1.91

-"

-1.91

POSITIVE LOGITS

高的

1.86

1.83

 throne

1.75

盪

1.73

езда

1.69

𝐓

1.68

婼

1.66

});

1.66

شعر

1.63

Activations Density 0.005%