INDEX

Explanations

variable assignments

The neuron fires most strongly on Java API type names (class identifiers) and related library‐call tokens.

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

 Перейти

-1.36

 keduanya

-1.27

 sebagainya

-1.26

 potongan

-1.24

 sogenannten

-1.22

 ponieważ

-1.21

 aforementioned

-1.20

 comenzará

-1.20

 kiedy

-1.20

you

-1.20

POSITIVE LOGITS

to

1.28

1.25

</h2>

1.25

 instead

1.24

the

1.22

each

1.19

 чём

1.18

we

1.16

 allt

1.16

 conquist

1.16

Activations Density 0.018%