INDEX

Explanations

describing components or functions

This neuron detects relation words used in formal definitions or descriptions—e.g. “part,” “role,” “houses,” “provides,” “involves”—that link a subject to its function or membership.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 their

-1.16

 which

-1.08

 there

-1.06

'].

-1.05

it

-1.02

or

-0.99

仍

-0.99

brtc

-0.97

voer

-0.96

'},

-0.96

POSITIVE LOGITS

 also

1.78

 myös

1.28

also

1.27

 TAMBÉM

1.23

もあり

1.23

 också

1.18

 inoltre

1.18

 також

1.16

 więc

1.16

 kuitenkin

1.15

Activations Density 0.192%