INDEX

Explanations

technical addresses and configurations

The neuron picks out programming‐related identifiers and file‐system path components (e.g. executable names, class or API tokens, directory names, etc.).

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

They

-0.81

⇡

-0.80

resten

-0.79

冠军

-0.74

Kalau

-0.74

 voltmeter

-0.73

GLP

-0.73

Naruto

-0.73

asyon

-0.72

כר

-0.72

POSITIVE LOGITS

 itself

1.00

its

0.99

as

0.91

ceus

0.91

 address

0.86

جميع

0.83

 siebie

0.83

oceros

0.81

Freq

0.78

 listing

0.78

Activations Density 0.090%