INDEX

Explanations

variable names and structure keys

The neuron fires on programming identifiers that end with an underscore (e.g. “car_”, “relay_”, “entry_”, “job_”), i.e. variable or column names with a trailing underscore.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

as

-2.02

人是

-1.60

his

-1.59

 last

-1.55

its

-1.53

now

-1.49

</sup>

-1.48

 mohou

-1.48

？】

-1.45

laints

-1.44

POSITIVE LOGITS

 Recuperado

1.53

辈子

1.47

1.45

场比赛

1.30

 chacune

1.26

1.24

gor

1.23

example

1.21

ogues

1.21

however

1.20

Activations Density 0.035%