INDEX

Explanations

enumeration and conjunctions

The neuron strongly activates on comma tokens (identifying commas in the text).

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

balleur

-1.29

hè

-1.27

 Algunas

-1.22

 ehemal

-1.20

を

-1.20

*),

-1.17

𝔲

-1.16

蜴

-1.16

が

-1.16

POSITIVE LOGITS

or

2.41

and

1.93

1.83

なんですね

1.48

なんですよ

1.38

}$.

1.37

 There

1.36

 thus

1.34

 thereby

1.34

all

1.31

Activations Density 0.165%