INDEX

Explanations

require_relative

The neuron strongly fires on long runs of alphanumeric and symbol sequences typical of base64 or other encoded/cryptographic key blobs.

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

or

-1.82

and

-1.79

These

-1.77

we

-1.73

but

-1.72

This

-1.70

 minutter

-1.67

 efficient

-1.64

because

-1.60

 diligent

-1.59

POSITIVE LOGITS

蓠

2.13

 ocorre

1.75

 다른

1.75

一切都

1.61



1.59

 recebe

1.59

 wist

1.58

跸

1.57

 emprego

1.57

 montrent

1.56

Activations Density 0.004%