INDEX

Explanations

being forged, reduces, eggs

The neuron is acting as a high‐frequency word detector, strongly activating on very common tokens (especially function words such as “of,” “in,” “the,” “and,” “a”) and other frequent English words.

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

-2.28

 such

-1.86

our

-1.76

my

-1.65

 aguj

-1.52

 this

-1.51

 nhưng

-1.49

 aceit

-1.49

 another

-1.47

 different

-1.43

POSITIVE LOGITS

</i>

1.75

 حتی

1.53

龶

1.48

 вони

1.47

for

1.46

by

1.45

 Leur

1.43

 даже

1.42

 with

1.41

 его

1.41

Activations Density 0.105%