INDEX

Explanations

verification and policy

The neuron detects occurrences of the root “verif-,” i.e. words like verify, verification, verifying, verified, etc.

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

ৢ

-2.64

-2.55

end

-2.39

⎆

-2.27

ти

-2.27

？！”

-2.23

 pouvons

-2.17

骉

-2.17

на

-2.13

ي

-2.13

POSITIVE LOGITS

This

3.36

It

3.30

We

2.64

我这

2.61

нутрен

2.50

</h3>

2.44

↑↑↑</

2.41

Like

2.38

There

2.36

After

2.36

Activations Density 0.008%