INDEX

Explanations

dates with day 1

The neuron selectively activates on numeric day‐or‐year tokens in date expressions (e.g. the “19” in “November 19, 2017” or “1” in “August 1, 1972”).

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

-1.69

-1.41

-1.34

 gastar

-1.34

 acompañado

-1.31

 sufre

-1.25

 dotyczące

-1.23

 volverá

-1.20

 pasará

-1.17

keeps

-1.17

POSITIVE LOGITS

 that

1.43

 even

1.30

fficients

1.26

onatos

1.23

of

1.20

 воспользоваться

1.16

 recompensa

1.16

❫

1.16

ÚBL

1.16

 because

1.14

Activations Density 0.017%