INDEX

Explanations

words starting with Hal

The neuron strongly activates on any token beginning with “Hal,” i.e. words or identifiers that start with the substring “Hal.”

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

-2.45

 However

-2.39

NOSIS

-2.27

 které

-2.25

 apparent

-2.23

 搞笑

-2.23

 surprising

-2.22

 These

-2.20

And

-2.20

かもしれませんが

-2.19

POSITIVE LOGITS

ଝ

2.98



2.58

埽

2.56

UAWEI

2.44

就

2.39

2.33

⚋

2.31

唎

2.30

𓐍

2.27

秾

2.22

Activations Density 0.013%