INDEX

Explanations

parenthetical descriptions

The neuron strongly activates on journalistic reporting verbs—especially “says”—marking quoted or attributed speech.

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

However

-1.88

ではまた

-1.63

 being

-1.62

 additional

-1.59

Однако

-1.57

 noticeable

-1.55

 enormous

-1.54

 також

-1.52

 نیز

-1.51

on

-1.49

POSITIVE LOGITS

 they

1.66

ेशा

1.65

 REALLY

1.64

ޏ

1.63

kabau

1.61

稔

1.60

还不错

1.59

“

1.59

 antaranya

1.59

 stattfinden

1.57

Activations Density 0.007%