INDEX

Explanations

The neuron fires strongly on content‐bearing nouns and proper nouns (e.g. product names, service terms, organization and role titles).

core topic terms—salient nouns and domain-specific keywords that denote the main subject or concepts in informational or technical text.

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

étaient

0.42

Were

0.41

 گئے۔

0.41

 estaban

0.40

 بودند

0.40

 fueran

0.38

都没

0.38

 తెలిపారు

0.37

ходили

0.37

 ディ

0.36

POSITIVE LOGITS

 helps

1.16

 provides

1.10

 enables

1.10

 gives

1.06

 tends

1.05

 contributes

1.04

 becomes

1.03

 involves

1.02

 creates

1.02

 allows

1.01

Activations Density 2.105%