index access or assignment (negative activations)

_attribute_name

Auto-Interp Tools

Configuration

Prompts (Dashboard)

16,384 prompts, 256 tokens each

Dataset (Dashboard)

neuronpedia/python-code-simplified

Embeds

IFrame

Link

Negative Logits

-0.11

-0.07

-0.05

ex

-0.05

index

-0.03

_of

-0.03

ub

-0.03

ap

-0.02

Positive Logits

ctionary

0.03

dividual

0.03

dices

0.02

maining

0.02

stance

0.02

duct

0.02

splay

0.02

stination

0.02

µ

0.02

tail

0.02

Activations Density 68.386%