INDEX

Explanations

user roles and groups

The neuron fires on code identifiers and literals referring to roles (e.g. “roles,” “Role,” “RoleId,” etc.), i.e. role-related fields and variables in code.

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

 that

-1.67

 každého

-1.43

 mutta

-1.39

 الوطنيه

-1.26

',

-1.25

 Arki

-1.21

却又

-1.21

 ſet

-1.20

 jette

-1.18

quête

-1.17

POSITIVE LOGITS

 then

1.30

了下

1.25

 kilka

1.21

 attacco

1.21

 wojsk

1.20

Funcion

1.18

 там

1.15

 giacca

1.13

了一

1.12

できるので

1.11

Activations Density 0.025%