INDEX

Explanations

defining what and how

The neuron activates on question‐style headings or lead‐in phrases (e.g. “What is…?”, “How it works,” “Why…?”) that introduce explanatory 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

-1.46

</h2>

-1.31

-1.30

pezifische

-1.22

通过

-1.21

 juist

-1.20

 ARISING

-1.19

もある

-1.15

forderungen

-1.14

 också

-1.13

POSITIVE LOGITS

していますが

1.32

 this

1.27

its

1.26

护理

1.23

 important

1.23

any

1.22

しているので

1.22

 helfen

1.19

CÁ

1.16

vió

1.13

Activations Density 0.062%