This neuron activates on UI/navigation labels and footer elements (e.g. “Footer,” “Recommended links,” “Follow us,” “Services,” “Contact”) in web‐scraped text.

New Auto-Interp

Configuration

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.11.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-0.10

-0.09

-0.08

POSITIVE LOGITS

cbc

0.09

CCA

0.09

coc

0.09

0.08

 تص

0.08

GC

0.08

co

0.08

Cyc

0.08

Activations Density 0.118%