INDEX

Explanations

and followed by a clause

The neuron consistently fires on the coordinating conjunction “and.”

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

PENDAHULUAN

-1.22

昶

-1.20

 something

-1.16

any

-1.14

OG

-1.13

 sánh

-1.12

 السياس

-1.11

 sehingga

-1.11

ølge

-1.10

 their

-1.09

POSITIVE LOGITS

it

1.59

 felt

1.52

 found

1.51

 have

1.48

 feel

1.45

 didn

1.41

พบ

1.27

was

1.25

saw

1.24

 hoped

1.24

Activations Density 0.079%