INDEX

Explanations

subject followed by action verb

This neuron fires on the first major content word at the start of a sentence (i.e. the sentence-opening “head” token).

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

 signifikan

-1.35

 protokol

-1.32

 praktik

-1.24

 radikal

-1.20

and

-1.16

Kamera

-1.14

cuatro

-1.13

jawab

-1.12

 bakteri

-1.09

 romantis

-1.09

POSITIVE LOGITS

 further

1.63

 also

1.54

 again

1.52

 especially

1.24

 wiem

1.22

 briefly

1.21

especially

1.21

 believe

1.19

 weiteres

1.18

 then

1.15

Activations Density 0.013%