INDEX

Explanations

parentheses and keywords

The neuron fires on numeric literals 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

carus

-1.26

 более

-1.23

mogen

-1.13

たしか

-1.08

 capucha

-1.06

greja

-1.05

«

-1.05

 один

-1.03

 avec

-1.03

 mehrere

-1.02

POSITIVE LOGITS

 would

1.45

by

1.39

at

1.27

 delicioso

1.27

端末

1.25

 hommage

1.18

but

1.18

 there

1.16

 manteau

1.16

Projekt

1.15

Activations Density 0.002%