INDEX

Explanations

regular expression syntax

The neuron fires on regex square‐bracket constructs (character‐class delimiters and their attached quantifiers).

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

-1.70

 during

-1.70

 because

-1.57

 when

-1.48

as

-1.45

 which

-1.42

 while

-1.41

for

-1.38

 through

-1.34

 became

-1.30

POSITIVE LOGITS

 revered

1.41

 discer

1.27

 tranquil

1.26

 strikingly

1.25

 quintessential

1.24

 gigantic

1.21

 astounding

1.21

Either

1.21

either

1.20

 majestic

1.20

Activations Density 0.012%