INDEX

Explanations

code snippets

The neuron is essentially firing on tokens that begin (and are indented into) nested code‐structure blocks—i.e. the opening delimiters and the first tokens of deeper, indented code sections.

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

 mereka

-1.50

augmenter

-1.48

 sowohl

-1.45

our

-1.43

 Wartawan

-1.41

 some

-1.35

 what

-1.34

᱑

-1.34

توضیحات

-1.32

at

-1.30

POSITIVE LOGITS

to

1.60



1.36

ORY

1.30

until

1.30

 другую

1.27

vê

1.26

 teto

1.23

 alho

1.22

 ferocious

1.20

≅

1.20

Activations Density 0.015%