INDEX

Explanations

come out

The neuron strongly fires on any form of the root “come” (come, came, coming) regardless of tense.

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

 để

-1.45

 that

-1.39

return

-1.33

medriver

-1.30

 untuk

-1.28

one

-1.27

></

-1.27

 zawsze

-1.23

用于

-1.23

<bos>

-1.20

POSITIVE LOGITS

 from

1.55

origami

1.45

龵

1.38

to

1.27

掬

1.26

excellents

1.23

in

1.20

wast

1.19

を含む

1.16

CHP

1.16

Activations Density 0.029%