INDEX

Explanations

specific contexts after certain tokens

This neuron lights up on substantive content words (nouns and verbs) rather than on common function words.

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

ほんとに

-1.45

ほんと

-1.38

 лікар

-1.35

calciatore

-1.27

色んな

-1.26

驚く

-1.26

あっという

-1.23

あぁ

-1.22

 lantaran

-1.22

 apunta

-1.20

POSITIVE LOGITS

 with

1.35

 نحو

1.13

幂

1.08

“

1.07

 numai

1.05

vă

1.03

use

1.02

ｾ

1.01

vâ

1.00

 edildi

1.00

Activations Density 0.003%