INDEX

Explanations

analyst roles and descriptors

The neuron activates on occurrences of the word “analyst.”

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

are

-3.66

You

-3.31

Its

-2.94

 algunos

-2.81

 Whether

-2.81

 подготовка

-2.73

 Therefore

-2.72

But

-2.70

 Denken

-2.58

 lección

-2.58

POSITIVE LOGITS

As

3.44

3.39

Not

3.36

3.22

From

3.14

3.09

3.06

2.94

ter

2.92

For

2.88

Activations Density 0.002%