INDEX

Explanations

reading instructions or documentation

The neuron detects instructional prompts—especially imperative verbs urging the reader to “read,” “check,” or otherwise consult documentation or guides.

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

-1.20

 only

-1.20

for

-1.16

in

-1.14

 mostly

-1.10

When

-1.09

 nothing

-1.09

There

-1.08

At

-1.06

 just

-1.05

POSITIVE LOGITS

the

2.41

our

1.38

 deuil

1.13

1.06

getPage

1.05

 vår

1.05

“:

1.05

 onze

1.05

iseur

1.05

corollary

1.03

Activations Density 0.045%