INDEX

Explanations

Control, profile, and status indicators

The neuron flags content-bearing tokens (nouns, verbs, adjectives, etc.)—i.e. it ignores common function words and lights up on semantically significant 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

 Setelah

-1.55

 Pemain

-1.33



-1.31

 contribue

-1.30

 mariposas

-1.28

じゃなくて

-1.27

(),

-1.27

 mencegah

-1.26

 Lokasi

-1.26

_),

-1.25

POSITIVE LOGITS

but

2.47

 when

1.95

 because

1.87

who

1.84

 although

1.64

 where

1.60

 потому

1.46

But

1.41

1.40

 quando

1.36

Activations Density 0.148%