INDEX

Explanations

phrases ending in -based, -derived, or -induced

The neuron strongly responds to the word “based,” particularly in hyphenated or compound-adjective contexts (e.g. “X-based”).

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

詆

-1.58

鸰

-1.35

 destacó

-1.31

 radiate

-1.25

Bukan

-1.24

嚶

-1.23

想了

-1.22

</b>

-1.22

afc

-1.22

がっ

-1.21

POSITIVE LOGITS

the

1.51

her

1.30

“

1.26

for

1.23

 milik

1.21

 landen

1.21

my

1.19

 efforts

1.17

 vernünf

1.13

Activations Density 0.160%