INDEX

Explanations

-cases, pome, ic

The neuron consistently activates on the initial word‐piece segments of longer content words—i.e. the prefixes where a multi‐syllable or compound term begins.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 ambayo

-3.09

which

-2.75

 lakini

-2.47

 vegano

-2.39

-2.36

 chini

-2.34

 yetu

-2.31

 広場

-2.27

tem

-2.23

 ambao

-2.23

POSITIVE LOGITS

 This

2.48

閏

2.39

%;

2.38

反抗

2.30

֟

2.20

 takich

2.20

 ciertos

2.20

 Apparently

2.17

]));

2.16

 clásica

2.16

Activations Density 0.063%