INDEX

Explanations

structure descriptions

This neuron primarily picks up formal relative‐clause markers in technical writing—e.g. parentheses and “which.”

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

abinieri

-1.45

brtc

-1.45

𝜙

-1.38

ཫ

-1.38

ۂ

-1.36

atized

-1.33

пустить

-1.32

に関しては

-1.30

 aree

-1.28

葙

-1.28

POSITIVE LOGITS

its

1.78

 that

1.39

 интерьера

1.38

.*;

1.38

etsk

1.32

慵

1.29

asanjo

1.28

 изделий

1.27

rowsiness

1.26

将其

1.25

Activations Density 0.610%