INDEX

Explanations

The neuron fires on running‐text narrative—especially English prose describing or explaining code or asking questions (e.g. “In this script we demonstrate how to use …,” “I am trying to …”).

examples and code syntax

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

𝗼

-1.39

 then

-1.16

Herzlich

-1.15

 правительства

-1.15

 анализ

-1.15

vs

-1.14

𝗺

-1.14

-1.13

 бил

-1.13

 commerces

-1.13

POSITIVE LOGITS

亟

1.59

instancetype

1.59

 collaborations

1.45

瞩

1.45

jedno

1.41

咂

1.39

ينج

1.37

 enthou

1.37

 samarbeid

1.37

 poë

1.34

Activations Density 0.377%