INDEX

Explanations

places, technologies, or measurements

The neuron consistently lights up on proper nouns and other named‐entity tokens—place names, organizations, framework and type names, etc.

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

 always

-1.28

 usually

-1.20

 всегда

-1.19

 thường

-1.08

always

-1.04

 siempre

-1.04

 toujours

-1.03

 תמיד

-1.02

 همیشه

-1.01

𝑸

-0.99

POSITIVE LOGITS

 would

1.27

 here

1.20

 will

1.12

 Would

1.03

Would

0.90

 herein

0.85

य

0.82

now

0.82

CollectionUtils

0.79

 było

0.79

Activations Density 0.159%