INDEX

Explanations

javascript require statements

The neuron is primarily picking up on numeric literals in code (i.e. standalone number 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

 gdyż

-1.38

 CGRectMake

-1.38

You

-1.35

"""

-1.35

्री

-1.34

Token

-1.33

ními

-1.32

ígenes

-1.31

AS

-1.30

POSITIVE LOGITS

 from

1.87

for

1.75

 between

1.43

on

1.43

 다른

1.40

 same

1.37

 նաև

1.34

 They

1.33

VPB

1.33

Theres

1.30

Activations Density 0.033%