INDEX

Explanations

old or new code

The neuron fires on words that name modes, versions, or configuration options (e.g. “old,” “new,” “default,” “OLTP,” “equivalent”).

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

and

-1.45

 itself

-1.34

不同

-1.32

čky

-1.27

astră

-1.23

↵↵

-1.21

 both

-1.20

uyasha

-1.20

׳

-1.20

 zarówno

-1.18

POSITIVE LOGITS

 conç

1.19

<i>

1.16

 rather

1.16

rather

1.14

instead

1.13

彼らの

1.11

 кто

1.11

still

1.04

يتها

1.04

ままで

1.03

Activations Density 0.295%