INDEX

Explanations

leads to

The neuron fires on causal “lead/leading to” constructions that set up a cause‐and‐effect relationship.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

lou

1.36

 dyads

1.34

ات

1.33

นี้

1.33

April

1.31

lays

1.30

๐

1.30

lal

1.28

 Così

1.27

that

1.26

POSITIVE LOGITS

可以

1.31

PERATURE

1.27

િક

1.25

ούς

1.25

ﺨ

1.24

 conclusion

1.22

 серьез

1.22

ul

1.20

ﺒ

1.20

цы

1.19

Activations Density 0.053%