INDEX

Explanations

sequences, transactions, technical terms

The neuron activates on technical or code-specific identifier tokens (e.g. class/type names, function names, and other domain-specific jargon).

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

-1.61

 côtes

-1.38

first

-1.37

vertreter

-1.33

-1.31

in

-1.30

 banques

-1.29

 بعضی

-1.27

POSITIVE LOGITS

to

1.53

so

1.45

珖

1.42

 four

1.41

 that

1.39

pompa

1.38

ёр

1.37

 three

1.32

bague

1.27

 Collaborative

1.26

Activations Density 0.014%