INDEX

Explanations

title or role followed by name

The neuron strongly activates on named individuals (proper names) together with their titles or roles in speaker‐attribution contexts.

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

各种

-1.70

 både

-1.48

 gainera

-1.46

 robar

-1.43

 огром

-1.43

詬

-1.37

Ingredientes

-1.36

 encender

-1.34

 באמצ

-1.34

 colgar

-1.34

POSITIVE LOGITS

of

2.73

 this

1.95

 that

1.81

at

1.59

 from

1.57

new

1.52

now

1.47

 solely

1.45

 after

1.41

 hoping

1.38

Activations Density 0.087%