INDEX

Explanations

after punctuation

The neuron fires on high‐information content words (i.e. open‐class nouns, verbs, and adjectives) rather than on common function words.

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

 albeit

1.23

but

1.22

 несмотря

1.16

 although

1.12

 mutta

1.10

 což

1.09

 którym

1.07

 lakini

1.05

 लेकिन

1.03

 जिसके

1.02

POSITIVE LOGITS

、『

0.93

、

0.84

Và

0.79

0.75

無い

0.74

『

0.71

")).

0.70

が無い

0.69

【

0.69

And

0.66

Activations Density 0.237%