INDEX

Explanations

code, programming concepts

The neuron fires most strongly on semantically heavy content words—especially key nouns and technical terms that carry the main topics of a passage.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 любят

0.63

원은

0.63

장은

0.58

 আছেন

0.58

率は

0.57

entes

0.57

 আছে

0.57

 используют

0.57

 দিয়েছে

0.57

 አለበት

0.55

POSITIVE LOGITS

 να

2.02

 להיות

1.94

ToBe

1.84

to

1.79

ที่จะ

1.62

 להת

1.53

să

1.31

ທີ່ຈະ

1.30

 ausz

1.29

abz

1.21

Activations Density 0.642%