INDEX

Explanations

naming specific entities and their attributes

The neuron activates on prominent multi-syllable nouns and proper names—especially technical terms, product or method names, and other domain-specific nouns.

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

 écouter

-1.32

 devast

-1.32

 remarkably

-1.30

 REALLY

-1.29

だけでなく

-1.27

ï

-1.27

 actuels

-1.26

 unveils

-1.23

他の

-1.23

LITER

-1.23

POSITIVE LOGITS

 that

1.61

 this

1.59

 which

1.47

and

1.30

 what

1.30

 after

1.30

 After

1.28

on

1.23

 whose

1.22

an

1.21

Activations Density 0.385%