INDEX

Explanations

verb followed by other words

The neuron flags auxiliary‐verb sequences forming perfect‐tense constructions (e.g. “has had,” “hasn’t sold,” “has been,” “had to wait”).

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

 Songtext

-1.23

 jurk

-1.17

古い

-1.16

 نداره

-1.14

 nudo

-1.14

ल्स

-1.13

Menurutnya

-1.12

zwi

-1.11

 bershka

-1.11

 eivät

-1.09

POSITIVE LOGITS

 being

1.60

to

1.34

on

1.30

mó

1.18

 ибо

1.16

侨

1.16

蒞

1.16

has

1.14

ultades

1.13

our

1.09

Activations Density 0.063%