INDEX

Explanations

Kul and Maha names

The neuron responds to the character sequence “kul” in words (e.g. in names like Kulkarni, Kulbin, or kul’te); it flags any token containing “kul.”

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

-2.48



-2.44

).

-2.42

%"

-2.30

?"

-2.25

⢌

-2.19

."

-2.16

 spur

-2.09

-2.08

VIATIONS

-2.05

POSITIVE LOGITS

衕

2.55

鯷

2.45

voegen

2.34

 verdaderamente

2.34

 vérifier

2.28

蛟

2.27

 abbin

2.25

 imagined

2.25

та

2.23

 asomb

2.23

Activations Density 0.003%