INDEX

Explanations

non-English words

The neuron strongly activates on unusually long tokens—especially multi‐syllabic or technical 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

ようです

-1.48

 should

-1.25

—

-1.24

-1.23

—

-1.22

 inclusief

-1.22

ようですが

-1.21

 five

-1.20

 this

-1.20

尔

-1.20

POSITIVE LOGITS

 kapit

1.37

や

1.36

 设计

1.35

karena

1.31

を

1.30

ється

1.30

schöne

1.28

 pregă

1.25

そして

1.24

汤

1.24

Activations Density 0.005%