INDEX

Explanations

compound technical terms

The neuron fires on unusually long tokens (e.g. technical identifiers or multi-syllable words) that exceed a length threshold.

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

 myriad

-1.31

ことも

-1.24

ཿ

-1.22

 unveils

-1.21



-1.20

 acht

-1.16

łoży

-1.15

 insanely

-1.13

 gwia

-1.12

྄

-1.12

POSITIVE LOGITS

 with

1.41

in

1.38

as

1.29

at

1.27

 perhaps

1.26

 Although

1.23

 mainly

1.21

 although

1.20

<bos>

1.19

 mostly

1.16

Activations Density 0.179%