INDEX

Explanations

phrase following punctuation

The neuron strongly activates on punctuation marks used as list separators—especially commas and semicolons.

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

 فبراير

-1.59

 utilisé

-1.52

หรือ

-1.49

 utilisée

-1.48

tichetta

-1.42

 aceste

-1.38

枚目

-1.38

隹

-1.36

劭

-1.34

 või

-1.31

POSITIVE LOGITS

2.06

an

1.77

and

1.66

 have

1.41

 Maintenant

1.32

//!

1.31

How

1.30

 jums

1.29

Zdroj

1.23

↵↵↵↵↵↵↵↵↵

1.21

Activations Density 0.297%