INDEX

Explanations

results showed

The neuron strongly activates on numeric tokens (e.g. section numbers, reference numbers, counts), i.e. it detects numbers.

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

for

-1.42

one

-1.34

Shouldn

-1.33

-1.32

all

-1.29

and

-1.23

 very

-1.20

 очень

-1.20

 full

-1.20

or

-1.18

POSITIVE LOGITS

 doigts

1.31

 raiſ

1.22

 Ministre

1.21

 そば

1.21

 digitale

1.19

 Ahli

1.16

 inond

1.16

 MONTE

1.16

 maniere

1.14

 reager

1.14

Activations Density 0.101%