INDEX

Explanations

speaks like a statesman:

This neuron flags prominent content words—particularly action verbs (e.g. “did,” “act,” “like”) and salient nouns or names (e.g. “Norma,” “Desmond,” “today”).

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

and

-1.14

 dezelfde

-0.96

ѹ

-0.93

 alojamientos

-0.92

ctiva

-0.91

 however

-0.91

entur

-0.90

 vervol

-0.88

 voldoende

-0.88

鏄

-0.88

POSITIVE LOGITS

1.88

1.48

–

1.28

 یعنی

1.23

：……

1.22

 yakni

1.22

—

1.20

—

1.13

ndash

1.09

--

1.02

Activations Density 0.075%