INDEX

Explanations

Go code structure keywords

The neuron fires most strongly on comment markers (“//”) and other non-code tokens like import path strings and doc comments—i.e. it identifies comment text (documentation) in source code.

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

-1.62

 caminhão

-1.30

 verlangt

-1.29

one

-1.28

 relógio

-1.27

Ꭳ

-1.27

 pregunto

-1.23

 unangemessene

-1.23

ilt

-1.23

 schlimm

-1.23

POSITIVE LOGITS

and

1.77

铵

1.43

’,

1.38

蚱

1.33

nyata

1.32

ށ

1.32

 milyon

1.31

ITÉ

1.30

 respirar

1.30

 şey

1.29

Activations Density 0.167%