INDEX

Explanations

plans discussed

The neuron tends to light up on longer, multi‐syllable content words—i.e. tokens of substantial length rather than short function 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

attore

-1.88

 الشاشة

-1.80

alucía

-1.63

引用元

-1.63

えっ

-1.60

 Gelegenheit

-1.57

 Gastgeber

-1.56

𝜌

-1.54

 стаття

-1.52

Etimología

-1.52

POSITIVE LOGITS

it

2.44

 another

1.75

 would

1.63

 without

1.49

 this

1.31

to

1.29

but

1.24

 immediately

1.24

if

1.24

 only

1.23

Activations Density 0.207%