INDEX

Explanations

of nouns and adjectives

The neuron fires on expository or instructional language—terms and markers you’d find in formal explanations or “how‐to” style text.

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

on

-1.48

 універ

-1.32

to

-1.25

 with

-1.24

 ketua

-1.23

 Originally

-1.22

 Despite

-1.21

Jokes

-1.18

 Anyways

-1.18

by

-1.16

POSITIVE LOGITS

౮

1.40

 新作

1.33

౧

1.29

 zoude

1.28

 drog

1.26

僦

1.26

ఽ

1.22

 первое

1.22

 niin

1.21

 then

1.20

Activations Density 0.647%