INDEX

Explanations

future actions or outcomes

The neuron flags the common introductory words and phrases that lead off an answer’s solution sentence (e.g. “This …,” “Gives me …,” “This creates …”).

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

 this

-1.69

 their

-1.62

 before

-1.55

 Every

-1.46

 they

-1.30

his

-1.29

 after

-1.28

 During

-1.27

 there

-1.24

 There

-1.23

POSITIVE LOGITS

 MISSION

1.34

 turco

1.34

 westfalen

1.30

 traktor

1.30

 DESK

1.28

 prodi

1.27

 gita

1.25

 CURRENT

1.24

 itali

1.24

ſi

1.23

Activations Density 0.152%